Your Gateway to Power, Energy, Datacenters, Bitcoin and AI

Dive into the latest industry updates, our exclusive Paperboy Newsletter, and curated insights designed to keep you informed. Stay ahead with minimal time spent.

Discover What Matters Most to You

Explore ONMINE’s curated content, from our Paperboy Newsletter to industry-specific insights tailored for energy, Bitcoin mining, and AI professionals.

AI

Lorem Ipsum is simply dummy text of the printing and typesetting industry.

Bitcoin:

Lorem Ipsum is simply dummy text of the printing and typesetting industry.

Datacenter:

Lorem Ipsum is simply dummy text of the printing and typesetting industry.

Energy:

Lorem Ipsum is simply dummy text of the printing and typesetting industry.

Shape
Discover What Matter Most to You

Featured Articles

Energy Secretary Keeps Critical Generation Available in Mid-Atlantic

WASHINGTON—U.S. Secretary of Energy Chris Wright today issued an emergency order to address critical grid reliability issues facing the mid-Atlantic region of the United States. The order directs PJM Interconnection L.L.C. (PJM), in coordination with Constellation Energy Corporation (Constellation Energy), to ensure Units 3 and 4 of the Eddystone Generating Station in Pennsylvania remain available to operate and to employ economic dispatch to minimize costs for the American people. The units were originally slated to shut down on May 31, 2025.   “The energy sources that perform when you need them most are the most valuable,” said U.S. Secretary of Energy Chris Wright. “President Trump and the Energy Department are committed to keeping critical generation available when demand is highest, reducing the risk of blackouts and ensuring Americans have affordable, reliable, and secure power—regardless of whether the wind is blowing or the sun is shining.”  Secretary Wright first ordered that the two Eddystone Generating Station units remain online past their planned retirement date in a May 30, 2025 emergency order. Subsequent orders were issued later in 2025 and 2026. Keeping these units operational has strengthened energy security in the PJM region, ensuring reliable electricity for American families and businesses. This order is in effect beginning on November 21, 2026, through February 18, 2027. 

Read More »

PHP2Shell aftershock: 18 more PHP flaws put enterprise web admins on alert

While there may be heavy focus on the 9.2-rated CVE-2026-129289, the lesser flaws still matter, Dickson said. For instance, in one case, one capital letter turned off encryption; a URL written as HTTPS:// or FTPS:// with an uppercase S could send passwords, tokens, form data, and files in plaintext, even through a redirect, to anyone watching the network. Not quite PHP2Shell level, but still high priority An attacker who exploits a PHP code-execution flaw gains access to whatever the PHP worker can touch, “which is just about everything,” Dickson said. Successful exploitation executes code with the privileges of the PHP worker, which holds the database password, API keys, cloud credentials, and session data. For example, he described what he called an “ugly” attack path:

Read More »

F5 scrambles to fix more critical BIG-IP security flaws

CVE-2026-121805: When an iRule that processes client traffic using the PEM:subscriber command is configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-120292: When an HTTP profile and an iRule using the MR:message route command are configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-120285: When an HTTP profile and an iRule containing the PSC:ip_address command are configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-116953: When an iRule that processes client traffic using the GENERICMESSAGE:route command is configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-123637: When BIG-IP APM is provisioned and a per-request policy that includes the Client Information agent and Endpoint Security agents is applied to a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE). The company said there are no mitigations for the flaws and advised customers to install the hotfixes on vulnerable devices, and to upgrade vulnerable versions without a listed fix to a patched version. In the same batch of out of band patches, F5 also addressed CVE-2026-126822, a high severity vulnerability integer underflow in BIG-IP APM SAML that could also crash TMM and lead to denial of service. This is the second critical issue in two months for APM; September’s zero-day bug was already being exploited when it was patched. F5 did not indicate that any of the current batch is as yet being exploited but given

Read More »

Hebbia sees improvements in financial research with GPT-6 Astra

Before a banker can recommend an acquisition to their client, they need to make sense of financial statements, customer contracts and management presentations. Hebbia’s Matrix helps bring that research together, linking answers to the documents behind them.Hebbia tested OpenAI’s GPT‑6 Astra to improve two parts of that work: finding the sources a question requires and building a presentation that follows the banker’s brief.“We’ve been strong partners with OpenAI over the course of Hebbia’s lifetime, and the leaps GPT-6 Astra has made in terms of quality outputs and accuracy have truly been surprising. GPT-6 Astra shines compared to any other model we’ve tested before on these two fronts.”—Kavita Anand, Member of Technical Staff, Product, HebbiaFrom source documents to presentationsIt could work like this: a banker is helping their client assess a strategic acquisition of an industrial-software company, with roughly 2,000 files to review. What makes its revenue durable? How does it compare with peers? In Hebbia’s hypothetical example, the banker could load the data room into Matrix and examine customer concentration, contract terms and churn in cited cells. Matrix could bring in peer filings and industry news to help develop the deal thesis, then turn that research into a diligence summary and valuation deck in the firm’s template.In that workflow, GPT‑6 Astra selects documents or Matrix cells to support an answer, then turns the analysis into native PowerPoint slides that follow the user’s structure and guidelines.“Traceability is really important to our users because they’re basing millions of dollars of decisions on what the AI actually comes back with,” Kavita says. “Citations help with traceability as well as giving users confidence that the AI found the right answer.”In Hebbia’s tests, GPT‑6 Astra combined broader source coverage with highly relevant citations. It scored 13.6 percentage points above the best alternative tested on citing the required types of sources, while 99.5% of the excerpts it cited were relevant to the question.Hebbia tested a prerelease build in August-September 2026, using the same financial research cases and grading criteria for each model. Domain experts wrote or audited the criteria against source documents.“When you’re looking at a data room with thousands of documents, that’s extremely important. Even a single percentage-point gain can yield a lot of marginal benefit,” Kavita says.A presentation also has to look like work the banker would send to a client.“I don’t want to create a deck and send that to my partner that looks like AI made it,” Kavita says. “I want it to be made in my firm’s branding, and I want it to look like a human-made slide.”GPT‑6 Astra led on both following the brief and presentation quality. In Hebbia’s tests, it met 97.9% of the applicable presentation requirements, putting it 14.4 percentage points ahead of the best alternative tested on that measure. Other models traded off polish and adherence to the brief; GPT‑6 Astra ranked first on both.“In the past, we’ve seen models that really shine on one or the other of these dimensions where they’re really good at craft, but the instruction following and adhering to guidelines is not strong, or vice versa. GPT‑6 Astra is the first model we’ve seen where it’s number one on both of those fronts,” Kavita says.In Hebbia, users can open a citation to see the underlying document and highlighted passage; slide footnotes lead back to the source material, too. A review could be as specific as checking a customer-concentration figure against the source table and correcting the reporting period before the deck goes to the investment committee.Hebbia wants bankers to spend less time reformatting decks and checking sources, and more time assessing a deal. Kavita describes what a better first draft could make possible:“What’s amazing about GPT-6 Astra is its ability to create really precise outputs and artifacts that adhere to user guidelines. You can imagine the hours this will save professionals when you get a deck that is closer to your final output. The overnighters that you have to pull get reduced to a few minutes of reviewing GPT-6 Astra’s output.”—Kavita Anand, Member of Technical Staff, Product, HebbiaA future role for agentsKavita sees a future role for agents that monitor information in the background and surface signals for finance professionals to assess.“Imagine getting an important signal a few hours before your competitors,” she says. “You can get started thinking about it and see and process more opportunities that a human is not manually hunting. It can unlock a lot of value for our users.”Hebbia wants to build on these results to help finance professionals assess deals with research they can trace back to its sources.OpenAI

Read More »

ExxonMobil takes FID on Angola Block 15 extension project

The Block 15 Lifespan Extension Project, about 145 km off Angola’s northern coast, will utilize existing infrastructure to develop additional resources throughout the Kizomba Hub area. The investment decision completes project FIDs following approval of Kizomba C redevelopment in 2025. ANPG also noted a hydrocarbon discovery in the block at the Tchihumba West-01ST1 exploration well, about 350 km northwest of Luanda. The discovery is the 21st in the block and the second this year. The well, drilled by the Valaris DS-9 in about 1,400 m of water, encountered 65.4 m of sandstone reservoir with hydrocarbon presence and good petrophysical properties. ExxonMobil is operator at Block 15 (36%) with partners Azule Energy Exploration (ANG) Ltd. (24%), Azule Energy ANG BV (18%), Equinor Angola (12%), and Sonongol E&P (10%).

Read More »

Combatting adversarial distillation

As frontier AI models become more capable, safeguards play an increasingly important role in how they can be used responsibly. Adversarial distillation creates a path for bad actors to reproduce these capabilities without appropriate safeguards. An actor may use a frontier model’s outputs without authorization to train another model with similar capabilities, then deploy it without comparable protections. For instance, an adversary might distill a model’s cyber capabilities and use the resulting model to help carry out cyberattacks that the original model would have refused to assist with or that its safeguards would have blocked.We’re sharing how OpenAI detects and disrupts adversarial distillation, how we’ve strengthened protections for our models’ protected reasoning, and what we’ve learned that can help strengthen defenses across the industry.Our approach to adversarial distillation combines adversarial testing, model training, classifiers, and API-level protections with investigations into coordinated attempts to extract model capabilities. These updated safeguards align more closely with those we use for biology and cybersecurity.We place particular emphasis on preventing attackers from extracting our models’ protected reasoning: the intermediate steps a model generates internally while working toward an answer, known as its chain-of-thought. These traces can contain information the final answer leaves out, making them especially valuable training data for another model.We’ve strengthened our defenses against protected reasoning extraction in GPT‑6.2 Astra and GPT‑6.2 Sol. In an internal adaptive evaluation using strong automated red-teaming, we tested whether attackers could recover protected reasoning the model had generated while answering an earlier prompt. The average attack success rate fell from 54.0% against GPT‑6 Astra with its original safeguards to 0.2% against GPT‑6.2 Astra with its updated safeguards.Why distillation is a safety issueDistillation is a common machine-learning technique in which one model learns from data produced by another. For example, a model can learn from another model’s answers, its reasoning, or its judgments about which answers are better and why. Beyond intellectual property concerns, unauthorized distillation poses serious safety risks. Through adversarial distillation, a second model can learn advanced frontier capabilities without inheriting the safeguards that govern how the original model can use them.The concern is especially clear in dual-use domains like cybersecurity and biology. A model that helps defenders find vulnerabilities in software could also help an attacker exploit them; a model that assists with biological research could also help someone develop a harmful pathogen. If such capabilities were transferred to a model with weaker safeguards, it may provide assistance that the original model would have refused, or that the original safeguards would have blocked.The importance of protecting exact reasoning traces for cyber safetyOur models help users with coding and defensive cybersecurity tasks. A second model trained on many of those benign outputs may learn skills that also apply to offensive cybersecurity tasks. To better understand this, we study how distillation, especially involving reasoning traces, may transfer risky cyber capabilities.In this setting, the model that produces the training data is called the teacher, and the model that learns from it is called the student. In an internal study on how cyber capabilities transfer, we used the content normally visible to users—final responses, reasoning summaries, and tool interactions—as our baseline for distillation, and tested how much additional cyber capability transfers when the student has access to the teacher’s exact reasoning traces.We started each training run from the same pretrained student model. We created a supervised fine-tuning dataset consisting primarily of trajectories from Astra-level teacher models working on coding and defensive cybersecurity tasks. Each trajectory includes all of the steps a model takes as it works through a task, including any tool interactions and its final response. We evaluated each student after fine-tuning and again after a lightweight round of reinforcement learning in cybersecurity and coding environments.After supervised fine-tuning, the student trained on exact reasoning traces generally outperformed the student trained on reasoning summaries. After reinforcement learning, the exact-trace student still scored higher at the largest inference budget tested on both benchmarks shown, while the summary-trained student scored higher at lower budgets.1ExploitBenchSEC-bench ProOur cyber distillation results suggest that protecting reasoning traces can limit cyber capability transfer, even when the underlying task is one allowed under our usage policies. Across many such interactions, exact reasoning traces can teach a student skills that generalize beyond the tasks the teacher originally performed. We will continue studying how much additional capability transfers across different student models and distillation training recipes.Defenses that reinforce one anotherWe defend against adversarial distillation at two complementary levels of our safeguard stack. At the request level, model behavior, classifier checks, and API controls limit reasoning trace extraction. At the campaign level, we look for patterns in request-level and account-level signals that can reveal coordinated extraction attempts. In turn, attacks uncovered in investigations become examples for training and testing our safeguards.Our defenses against adversarial distillation combine safeguards for individual requests with campaign detection and enforcement. Findings from investigations and automated red-teaming help strengthen those safeguards.Protecting individual requests and responsesDefensive training for model behaviorAdversarial training teaches models to recognize reasoning extraction attempts and refuse to disclose protected reasoning, while preserving responses to legitimate requests for explanations and assistance. Adversarial training examples come from successful attacks identified through internal automated red-teaming, investigations, and responsible disclosures of potential security weaknesses from outside researchers or industry partners.Classifier safeguardsWe use input classifiers to detect attempts to extract protected reasoning, and output classifiers to detect leaks of protected reasoning in assistant messages and tool-call arguments. These safeguards apply to both streaming and non-streaming responses.Tool-call coverage is important because a model can expose protected reasoning in a tool’s arguments without including it in its final answer. Output classifiers can block suspected leaks before that content reaches the caller.API-level protectionsAPI-level protections restrict how protected reasoning can be reused across interactions and models. Our September disclosure described attacks that moved encrypted reasoning from one conversation into another and asked the model to reveal the reasoning in user-readable form. We strengthened protections across user, workspace, organization, and model-family boundaries, including closing a pathway that could expose another user’s encrypted reasoning.With GPT‑6.2 Astra and GPT‑6.2 Sol, we tie a model’s protected reasoning to the conversation history that produced it. As new messages are added, the model can build on its earlier reasoning. If an earlier message is rewritten, our API drops the reasoning that followed it⁠(opens in a new window). We also prevent older model families with weaker safeguards from reading GPT‑6.2 Astra and Sol’s protected reasoning.Investigating and disrupting coordinated campaignsWhen our systems detect unauthorized adversarial distillation campaigns, which often include coordinated attempts to extract our models’ reasoning, we investigate the accounts responsible. When we identify violations, we enforce against the network of accounts, consistent with our policies and terms of service. Through investigations of each campaign, we generate insights that allow us to upgrade our defenses in order to better detect similar activity. We also share indicators of compromise with other model providers so they can investigate related abuse on their own services.This work extends beyond our own platform. Our consistent-standards policy establishes shared standards for safeguards and coordinated enforcement across providers offering our models. We work closely with partners where appropriate to uphold these standards and strengthen protections across deployments.Sharing relevant findings across industry and government where appropriate can help disrupt campaigns and enable model developers and providers to better identify and defend against attempts to distill frontier models.Measuring progressResistance to reasoning extractionTo measure how well our request-level safeguards work together, we test whether attackers can still extract protected reasoning in a setup that closely mirrors production deployments of our models and safeguards. In static evaluations, we replay a fixed set of previously successful attacks to see which still expose protected reasoning. In adaptive evaluations, automated AI red-teamers adjust their approach based on how the system responds, which can uncover weaknesses the static tests miss.We also vary what testers can see. Black-box testing gives them the same view an external user would have, while white-box testing gives them additional information about the system to help them develop stronger attacks. In our adaptive black-box evaluation, we tested whether attackers could recover protected reasoning the model had generated while answering an earlier request or the current one. On GPT‑6.2 Astra with its updated safeguards, the average attack success rate was below 1% for each attack type. Both rates were substantially lower than on GPT‑6 Astra with its original safeguards. We’re now applying selected safeguard improvements to GPT‑6 Astra as well.Attack success rate (ASR) in adaptive black-box red-teaming against GPT‑6 Astra and GPT‑6.2 Astra, aggregated across six attacker models. Lower ASR is better.Campaign detection and enforcementWe have seen multiple unauthorized adversarial distillation campaigns on our platform this year, and shared initial findings in our September report. Our weekly enforcement actions against these campaigns have ranged in scale from thousands to tens of thousands of accounts, with the rate increasing significantly this past summer.In order to measure the effectiveness of our efforts, we also track the prevalence of distillation abuse, as well as the time between detection and enforcement. Our objective is to constrain the volume and value of the data distillers are able to obtain, to increase the cost of distillation campaigns, and to shorten the time needed to respond to new attacks. We continue to improve these defenses through adversarial testing and investigations.The national security implications of adversarial distillationAdversarial distillation also poses a national security risk. A foreign actor that reproduces advanced capabilities from a U.S. model can deploy them without the protections its developer put in place. In February, we briefed⁠(opens in a new window) Congress on activity indicating that DeepSeek was attempting to distill U.S. frontier models. In September, we reported disrupting a campaign to extract our models’ protected reasoning and attributed a core part of that activity to individuals associated with Moonshot AI, the developer of Kimi. We plan to share more soon about recent campaigns we’ve disrupted.We support legislation that would give the U.S. government more tools to identify and sanction foreign actors that illicitly extract capabilities from American AI models. We’ve also told Congress we support the Chip Security Act, the AI OVERWATCH Act, and the MATCH Act. Stronger export controls can limit adversaries’ access to the hardware needed to train frontier models. Distillation defenses make it harder to acquire some of the same capabilities from models already trained. Both efforts help maintain a U.S. lead in frontier AI, giving developers more room to slow a release when they need time to test safeguards.How others can build on this workSupporting the safe proliferation of frontier capabilities is an important goal of our anti-distillation work. Our experience so far suggests several useful steps that other model developers and platforms can take:Protect private reasoning across the full interaction. Check assistant messages and tool-call arguments for leaks, and restrict how encrypted reasoning can be reused across conversations and models.Keep successful extraction attack logs found through red-teaming, production monitoring, or responsible disclosure as regression tests. Where appropriate, and consistent with applicable privacy policies, use those examples to improve model training and classifiers. When evaluating updated defenses, replay the original attacks and test whether an adaptive red-teamer can still find a way through those defenses.Combine request-level mitigations with campaign-level investigation and enforcement.Coordinate safeguard requirements and enforcement with deployment partners, and share expertise with other AI model developers, to ensure a high floor of distillation safeguards across the industry.We have previously also shared findings through the Frontier Model Forum⁠(opens in a new window) and appropriate government information-sharing channels to help other developers strengthen their defenses while keeping the details attackers could exploit private. We’ll continue working with researchers, developers, and partners to share what we learn about emerging attacks and effective defenses.As we make frontier AI broadly available, our defenses against adversarial distillation aim to reduce the risk that its capabilities are reproduced without adequate safeguards, supporting our mission to ensure that artificial general intelligence is deployed safely and benefits all of humanity.

Read More »

Energy Secretary Keeps Critical Generation Available in Mid-Atlantic

WASHINGTON—U.S. Secretary of Energy Chris Wright today issued an emergency order to address critical grid reliability issues facing the mid-Atlantic region of the United States. The order directs PJM Interconnection L.L.C. (PJM), in coordination with Constellation Energy Corporation (Constellation Energy), to ensure Units 3 and 4 of the Eddystone Generating Station in Pennsylvania remain available to operate and to employ economic dispatch to minimize costs for the American people. The units were originally slated to shut down on May 31, 2025.   “The energy sources that perform when you need them most are the most valuable,” said U.S. Secretary of Energy Chris Wright. “President Trump and the Energy Department are committed to keeping critical generation available when demand is highest, reducing the risk of blackouts and ensuring Americans have affordable, reliable, and secure power—regardless of whether the wind is blowing or the sun is shining.”  Secretary Wright first ordered that the two Eddystone Generating Station units remain online past their planned retirement date in a May 30, 2025 emergency order. Subsequent orders were issued later in 2025 and 2026. Keeping these units operational has strengthened energy security in the PJM region, ensuring reliable electricity for American families and businesses. This order is in effect beginning on November 21, 2026, through February 18, 2027. 

Read More »

PHP2Shell aftershock: 18 more PHP flaws put enterprise web admins on alert

While there may be heavy focus on the 9.2-rated CVE-2026-129289, the lesser flaws still matter, Dickson said. For instance, in one case, one capital letter turned off encryption; a URL written as HTTPS:// or FTPS:// with an uppercase S could send passwords, tokens, form data, and files in plaintext, even through a redirect, to anyone watching the network. Not quite PHP2Shell level, but still high priority An attacker who exploits a PHP code-execution flaw gains access to whatever the PHP worker can touch, “which is just about everything,” Dickson said. Successful exploitation executes code with the privileges of the PHP worker, which holds the database password, API keys, cloud credentials, and session data. For example, he described what he called an “ugly” attack path:

Read More »

F5 scrambles to fix more critical BIG-IP security flaws

CVE-2026-121805: When an iRule that processes client traffic using the PEM:subscriber command is configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-120292: When an HTTP profile and an iRule using the MR:message route command are configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-120285: When an HTTP profile and an iRule containing the PSC:ip_address command are configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-116953: When an iRule that processes client traffic using the GENERICMESSAGE:route command is configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-123637: When BIG-IP APM is provisioned and a per-request policy that includes the Client Information agent and Endpoint Security agents is applied to a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE). The company said there are no mitigations for the flaws and advised customers to install the hotfixes on vulnerable devices, and to upgrade vulnerable versions without a listed fix to a patched version. In the same batch of out of band patches, F5 also addressed CVE-2026-126822, a high severity vulnerability integer underflow in BIG-IP APM SAML that could also crash TMM and lead to denial of service. This is the second critical issue in two months for APM; September’s zero-day bug was already being exploited when it was patched. F5 did not indicate that any of the current batch is as yet being exploited but given

Read More »

Hebbia sees improvements in financial research with GPT-6 Astra

Before a banker can recommend an acquisition to their client, they need to make sense of financial statements, customer contracts and management presentations. Hebbia’s Matrix helps bring that research together, linking answers to the documents behind them.Hebbia tested OpenAI’s GPT‑6 Astra to improve two parts of that work: finding the sources a question requires and building a presentation that follows the banker’s brief.“We’ve been strong partners with OpenAI over the course of Hebbia’s lifetime, and the leaps GPT-6 Astra has made in terms of quality outputs and accuracy have truly been surprising. GPT-6 Astra shines compared to any other model we’ve tested before on these two fronts.”—Kavita Anand, Member of Technical Staff, Product, HebbiaFrom source documents to presentationsIt could work like this: a banker is helping their client assess a strategic acquisition of an industrial-software company, with roughly 2,000 files to review. What makes its revenue durable? How does it compare with peers? In Hebbia’s hypothetical example, the banker could load the data room into Matrix and examine customer concentration, contract terms and churn in cited cells. Matrix could bring in peer filings and industry news to help develop the deal thesis, then turn that research into a diligence summary and valuation deck in the firm’s template.In that workflow, GPT‑6 Astra selects documents or Matrix cells to support an answer, then turns the analysis into native PowerPoint slides that follow the user’s structure and guidelines.“Traceability is really important to our users because they’re basing millions of dollars of decisions on what the AI actually comes back with,” Kavita says. “Citations help with traceability as well as giving users confidence that the AI found the right answer.”In Hebbia’s tests, GPT‑6 Astra combined broader source coverage with highly relevant citations. It scored 13.6 percentage points above the best alternative tested on citing the required types of sources, while 99.5% of the excerpts it cited were relevant to the question.Hebbia tested a prerelease build in August-September 2026, using the same financial research cases and grading criteria for each model. Domain experts wrote or audited the criteria against source documents.“When you’re looking at a data room with thousands of documents, that’s extremely important. Even a single percentage-point gain can yield a lot of marginal benefit,” Kavita says.A presentation also has to look like work the banker would send to a client.“I don’t want to create a deck and send that to my partner that looks like AI made it,” Kavita says. “I want it to be made in my firm’s branding, and I want it to look like a human-made slide.”GPT‑6 Astra led on both following the brief and presentation quality. In Hebbia’s tests, it met 97.9% of the applicable presentation requirements, putting it 14.4 percentage points ahead of the best alternative tested on that measure. Other models traded off polish and adherence to the brief; GPT‑6 Astra ranked first on both.“In the past, we’ve seen models that really shine on one or the other of these dimensions where they’re really good at craft, but the instruction following and adhering to guidelines is not strong, or vice versa. GPT‑6 Astra is the first model we’ve seen where it’s number one on both of those fronts,” Kavita says.In Hebbia, users can open a citation to see the underlying document and highlighted passage; slide footnotes lead back to the source material, too. A review could be as specific as checking a customer-concentration figure against the source table and correcting the reporting period before the deck goes to the investment committee.Hebbia wants bankers to spend less time reformatting decks and checking sources, and more time assessing a deal. Kavita describes what a better first draft could make possible:“What’s amazing about GPT-6 Astra is its ability to create really precise outputs and artifacts that adhere to user guidelines. You can imagine the hours this will save professionals when you get a deck that is closer to your final output. The overnighters that you have to pull get reduced to a few minutes of reviewing GPT-6 Astra’s output.”—Kavita Anand, Member of Technical Staff, Product, HebbiaA future role for agentsKavita sees a future role for agents that monitor information in the background and surface signals for finance professionals to assess.“Imagine getting an important signal a few hours before your competitors,” she says. “You can get started thinking about it and see and process more opportunities that a human is not manually hunting. It can unlock a lot of value for our users.”Hebbia wants to build on these results to help finance professionals assess deals with research they can trace back to its sources.OpenAI

Read More »

ExxonMobil takes FID on Angola Block 15 extension project

The Block 15 Lifespan Extension Project, about 145 km off Angola’s northern coast, will utilize existing infrastructure to develop additional resources throughout the Kizomba Hub area. The investment decision completes project FIDs following approval of Kizomba C redevelopment in 2025. ANPG also noted a hydrocarbon discovery in the block at the Tchihumba West-01ST1 exploration well, about 350 km northwest of Luanda. The discovery is the 21st in the block and the second this year. The well, drilled by the Valaris DS-9 in about 1,400 m of water, encountered 65.4 m of sandstone reservoir with hydrocarbon presence and good petrophysical properties. ExxonMobil is operator at Block 15 (36%) with partners Azule Energy Exploration (ANG) Ltd. (24%), Azule Energy ANG BV (18%), Equinor Angola (12%), and Sonongol E&P (10%).

Read More »

Combatting adversarial distillation

As frontier AI models become more capable, safeguards play an increasingly important role in how they can be used responsibly. Adversarial distillation creates a path for bad actors to reproduce these capabilities without appropriate safeguards. An actor may use a frontier model’s outputs without authorization to train another model with similar capabilities, then deploy it without comparable protections. For instance, an adversary might distill a model’s cyber capabilities and use the resulting model to help carry out cyberattacks that the original model would have refused to assist with or that its safeguards would have blocked.We’re sharing how OpenAI detects and disrupts adversarial distillation, how we’ve strengthened protections for our models’ protected reasoning, and what we’ve learned that can help strengthen defenses across the industry.Our approach to adversarial distillation combines adversarial testing, model training, classifiers, and API-level protections with investigations into coordinated attempts to extract model capabilities. These updated safeguards align more closely with those we use for biology and cybersecurity.We place particular emphasis on preventing attackers from extracting our models’ protected reasoning: the intermediate steps a model generates internally while working toward an answer, known as its chain-of-thought. These traces can contain information the final answer leaves out, making them especially valuable training data for another model.We’ve strengthened our defenses against protected reasoning extraction in GPT‑6.2 Astra and GPT‑6.2 Sol. In an internal adaptive evaluation using strong automated red-teaming, we tested whether attackers could recover protected reasoning the model had generated while answering an earlier prompt. The average attack success rate fell from 54.0% against GPT‑6 Astra with its original safeguards to 0.2% against GPT‑6.2 Astra with its updated safeguards.Why distillation is a safety issueDistillation is a common machine-learning technique in which one model learns from data produced by another. For example, a model can learn from another model’s answers, its reasoning, or its judgments about which answers are better and why. Beyond intellectual property concerns, unauthorized distillation poses serious safety risks. Through adversarial distillation, a second model can learn advanced frontier capabilities without inheriting the safeguards that govern how the original model can use them.The concern is especially clear in dual-use domains like cybersecurity and biology. A model that helps defenders find vulnerabilities in software could also help an attacker exploit them; a model that assists with biological research could also help someone develop a harmful pathogen. If such capabilities were transferred to a model with weaker safeguards, it may provide assistance that the original model would have refused, or that the original safeguards would have blocked.The importance of protecting exact reasoning traces for cyber safetyOur models help users with coding and defensive cybersecurity tasks. A second model trained on many of those benign outputs may learn skills that also apply to offensive cybersecurity tasks. To better understand this, we study how distillation, especially involving reasoning traces, may transfer risky cyber capabilities.In this setting, the model that produces the training data is called the teacher, and the model that learns from it is called the student. In an internal study on how cyber capabilities transfer, we used the content normally visible to users—final responses, reasoning summaries, and tool interactions—as our baseline for distillation, and tested how much additional cyber capability transfers when the student has access to the teacher’s exact reasoning traces.We started each training run from the same pretrained student model. We created a supervised fine-tuning dataset consisting primarily of trajectories from Astra-level teacher models working on coding and defensive cybersecurity tasks. Each trajectory includes all of the steps a model takes as it works through a task, including any tool interactions and its final response. We evaluated each student after fine-tuning and again after a lightweight round of reinforcement learning in cybersecurity and coding environments.After supervised fine-tuning, the student trained on exact reasoning traces generally outperformed the student trained on reasoning summaries. After reinforcement learning, the exact-trace student still scored higher at the largest inference budget tested on both benchmarks shown, while the summary-trained student scored higher at lower budgets.1ExploitBenchSEC-bench ProOur cyber distillation results suggest that protecting reasoning traces can limit cyber capability transfer, even when the underlying task is one allowed under our usage policies. Across many such interactions, exact reasoning traces can teach a student skills that generalize beyond the tasks the teacher originally performed. We will continue studying how much additional capability transfers across different student models and distillation training recipes.Defenses that reinforce one anotherWe defend against adversarial distillation at two complementary levels of our safeguard stack. At the request level, model behavior, classifier checks, and API controls limit reasoning trace extraction. At the campaign level, we look for patterns in request-level and account-level signals that can reveal coordinated extraction attempts. In turn, attacks uncovered in investigations become examples for training and testing our safeguards.Our defenses against adversarial distillation combine safeguards for individual requests with campaign detection and enforcement. Findings from investigations and automated red-teaming help strengthen those safeguards.Protecting individual requests and responsesDefensive training for model behaviorAdversarial training teaches models to recognize reasoning extraction attempts and refuse to disclose protected reasoning, while preserving responses to legitimate requests for explanations and assistance. Adversarial training examples come from successful attacks identified through internal automated red-teaming, investigations, and responsible disclosures of potential security weaknesses from outside researchers or industry partners.Classifier safeguardsWe use input classifiers to detect attempts to extract protected reasoning, and output classifiers to detect leaks of protected reasoning in assistant messages and tool-call arguments. These safeguards apply to both streaming and non-streaming responses.Tool-call coverage is important because a model can expose protected reasoning in a tool’s arguments without including it in its final answer. Output classifiers can block suspected leaks before that content reaches the caller.API-level protectionsAPI-level protections restrict how protected reasoning can be reused across interactions and models. Our September disclosure described attacks that moved encrypted reasoning from one conversation into another and asked the model to reveal the reasoning in user-readable form. We strengthened protections across user, workspace, organization, and model-family boundaries, including closing a pathway that could expose another user’s encrypted reasoning.With GPT‑6.2 Astra and GPT‑6.2 Sol, we tie a model’s protected reasoning to the conversation history that produced it. As new messages are added, the model can build on its earlier reasoning. If an earlier message is rewritten, our API drops the reasoning that followed it⁠(opens in a new window). We also prevent older model families with weaker safeguards from reading GPT‑6.2 Astra and Sol’s protected reasoning.Investigating and disrupting coordinated campaignsWhen our systems detect unauthorized adversarial distillation campaigns, which often include coordinated attempts to extract our models’ reasoning, we investigate the accounts responsible. When we identify violations, we enforce against the network of accounts, consistent with our policies and terms of service. Through investigations of each campaign, we generate insights that allow us to upgrade our defenses in order to better detect similar activity. We also share indicators of compromise with other model providers so they can investigate related abuse on their own services.This work extends beyond our own platform. Our consistent-standards policy establishes shared standards for safeguards and coordinated enforcement across providers offering our models. We work closely with partners where appropriate to uphold these standards and strengthen protections across deployments.Sharing relevant findings across industry and government where appropriate can help disrupt campaigns and enable model developers and providers to better identify and defend against attempts to distill frontier models.Measuring progressResistance to reasoning extractionTo measure how well our request-level safeguards work together, we test whether attackers can still extract protected reasoning in a setup that closely mirrors production deployments of our models and safeguards. In static evaluations, we replay a fixed set of previously successful attacks to see which still expose protected reasoning. In adaptive evaluations, automated AI red-teamers adjust their approach based on how the system responds, which can uncover weaknesses the static tests miss.We also vary what testers can see. Black-box testing gives them the same view an external user would have, while white-box testing gives them additional information about the system to help them develop stronger attacks. In our adaptive black-box evaluation, we tested whether attackers could recover protected reasoning the model had generated while answering an earlier request or the current one. On GPT‑6.2 Astra with its updated safeguards, the average attack success rate was below 1% for each attack type. Both rates were substantially lower than on GPT‑6 Astra with its original safeguards. We’re now applying selected safeguard improvements to GPT‑6 Astra as well.Attack success rate (ASR) in adaptive black-box red-teaming against GPT‑6 Astra and GPT‑6.2 Astra, aggregated across six attacker models. Lower ASR is better.Campaign detection and enforcementWe have seen multiple unauthorized adversarial distillation campaigns on our platform this year, and shared initial findings in our September report. Our weekly enforcement actions against these campaigns have ranged in scale from thousands to tens of thousands of accounts, with the rate increasing significantly this past summer.In order to measure the effectiveness of our efforts, we also track the prevalence of distillation abuse, as well as the time between detection and enforcement. Our objective is to constrain the volume and value of the data distillers are able to obtain, to increase the cost of distillation campaigns, and to shorten the time needed to respond to new attacks. We continue to improve these defenses through adversarial testing and investigations.The national security implications of adversarial distillationAdversarial distillation also poses a national security risk. A foreign actor that reproduces advanced capabilities from a U.S. model can deploy them without the protections its developer put in place. In February, we briefed⁠(opens in a new window) Congress on activity indicating that DeepSeek was attempting to distill U.S. frontier models. In September, we reported disrupting a campaign to extract our models’ protected reasoning and attributed a core part of that activity to individuals associated with Moonshot AI, the developer of Kimi. We plan to share more soon about recent campaigns we’ve disrupted.We support legislation that would give the U.S. government more tools to identify and sanction foreign actors that illicitly extract capabilities from American AI models. We’ve also told Congress we support the Chip Security Act, the AI OVERWATCH Act, and the MATCH Act. Stronger export controls can limit adversaries’ access to the hardware needed to train frontier models. Distillation defenses make it harder to acquire some of the same capabilities from models already trained. Both efforts help maintain a U.S. lead in frontier AI, giving developers more room to slow a release when they need time to test safeguards.How others can build on this workSupporting the safe proliferation of frontier capabilities is an important goal of our anti-distillation work. Our experience so far suggests several useful steps that other model developers and platforms can take:Protect private reasoning across the full interaction. Check assistant messages and tool-call arguments for leaks, and restrict how encrypted reasoning can be reused across conversations and models.Keep successful extraction attack logs found through red-teaming, production monitoring, or responsible disclosure as regression tests. Where appropriate, and consistent with applicable privacy policies, use those examples to improve model training and classifiers. When evaluating updated defenses, replay the original attacks and test whether an adaptive red-teamer can still find a way through those defenses.Combine request-level mitigations with campaign-level investigation and enforcement.Coordinate safeguard requirements and enforcement with deployment partners, and share expertise with other AI model developers, to ensure a high floor of distillation safeguards across the industry.We have previously also shared findings through the Frontier Model Forum⁠(opens in a new window) and appropriate government information-sharing channels to help other developers strengthen their defenses while keeping the details attackers could exploit private. We’ll continue working with researchers, developers, and partners to share what we learn about emerging attacks and effective defenses.As we make frontier AI broadly available, our defenses against adversarial distillation aim to reduce the risk that its capabilities are reproduced without adequate safeguards, supporting our mission to ensure that artificial general intelligence is deployed safely and benefits all of humanity.

Read More »

Energy Secretary Keeps Critical Generation Available in Mid-Atlantic

WASHINGTON—U.S. Secretary of Energy Chris Wright today issued an emergency order to address critical grid reliability issues facing the mid-Atlantic region of the United States. The order directs PJM Interconnection L.L.C. (PJM), in coordination with Constellation Energy Corporation (Constellation Energy), to ensure Units 3 and 4 of the Eddystone Generating Station in Pennsylvania remain available to operate and to employ economic dispatch to minimize costs for the American people. The units were originally slated to shut down on May 31, 2025.   “The energy sources that perform when you need them most are the most valuable,” said U.S. Secretary of Energy Chris Wright. “President Trump and the Energy Department are committed to keeping critical generation available when demand is highest, reducing the risk of blackouts and ensuring Americans have affordable, reliable, and secure power—regardless of whether the wind is blowing or the sun is shining.”  Secretary Wright first ordered that the two Eddystone Generating Station units remain online past their planned retirement date in a May 30, 2025 emergency order. Subsequent orders were issued later in 2025 and 2026. Keeping these units operational has strengthened energy security in the PJM region, ensuring reliable electricity for American families and businesses. This order is in effect beginning on November 21, 2026, through February 18, 2027. 

Read More »

ExxonMobil takes FID on Angola Block 15 extension project

The Block 15 Lifespan Extension Project, about 145 km off Angola’s northern coast, will utilize existing infrastructure to develop additional resources throughout the Kizomba Hub area. The investment decision completes project FIDs following approval of Kizomba C redevelopment in 2025. ANPG also noted a hydrocarbon discovery in the block at the Tchihumba West-01ST1 exploration well, about 350 km northwest of Luanda. The discovery is the 21st in the block and the second this year. The well, drilled by the Valaris DS-9 in about 1,400 m of water, encountered 65.4 m of sandstone reservoir with hydrocarbon presence and good petrophysical properties. ExxonMobil is operator at Block 15 (36%) with partners Azule Energy Exploration (ANG) Ltd. (24%), Azule Energy ANG BV (18%), Equinor Angola (12%), and Sonongol E&P (10%).

Read More »

Market reforms bolster grid, may lower costs this winter: FERC

Listen to the article 4 min This audio is auto-generated. Please let us know if you have feedback. Electric market reforms in the West and Northeast may bolster grid reliability and, in some cases, lower costs this winter, according to the Federal Energy Regulatory Commission’s annual winter market and reliability report, released Thursday. Expanded markets help offset drought impacts In the West, the California Independent System Operator launched the Extended Day Ahead Market and the Southwest Power Pool expanded into the Western Interconnection earlier this year. “EDAM and SPP’s expansion should provide ratepayers in much of the West with production cost savings and improved reliability in winter 2026-2027, particularly during cold weather events,” FERC staff said in the report. This winter, the markets’ improved coordination of generation and reduced need for reserve power supplies will help offset lower power output from the Glen Canyon and Hoover dams due to historically low water levels in the Colorado River basin, staff said. “On a day-to-day basis, these conditions can create congestion and affect system operators’ ability to respond to voltage support challenges, event response mitigation needs, or any additional grid stressors,” staff said in the report. About 45% of the continental United States and 95% of the Intermountain West region is in a drought, according to the report. Hydro, oil pose challenges for Northeast In the Northeast, ISO New England lowered its capacity performance pay rate, which should lower electricity prices during emergency conditions or capacity performance events this winter, staff said. Also, the New York Independent System Operator changed its installed capacity market rules to reflect uncertainty about how much power Hydro–Québec will deliver over the 1,250-MW Champlain Hudson Power Express transmission line in the winter, they noted. The U.S. subsidiary of the Canadian utility Hydro-Québec — the primary reservation holder — has an obligation

Read More »

Dangote awards contracts advancing Lekki refining complex expansion

Dangote Group’s Dangote Petroleum Refinery and Petrochemicals FZE (DPRP) has awarded separate contracts to Kirloskar Group of India’s Kirloskar Brothers Ltd. (KBL) and Kirloskar Ebara Pumps Ltd. (KEPL)—a KBL joint venture with Ebara Corp. of Japan—for delivery of engineering services and equipment as part of a phased capacity expansion currently underway at DPRP’s 650,000-b/d refinery and petrochemical complex in the Lekki Free Zone near Lagos, Nigeria. In a Nov. 20 regulatory filing submitted to both BSE Ltd. and The National Stock Exchange of India Ltd., KBL said the combined value of the orders under the contracts amounted to about $38.94 million, $21.34 million of which would cover KEPL’s scope of work, with KBL to retain the remaining $17.60 million as compensation for a leg of the project. The service provider offered no additional details regarding either its own of KEPL’s scope of deliverables under the agreement other than that the work was to include engineering expertise and provision of “specialized pumping solutions for demanding energy applications.” The Lekki complex expansion involves increasing the refinery’s crude processing capacity by an additional 750,000 b/d to 1.4 million b/d in two phases to be executed between 2025-28 and 2028-30, respectively. The proposed expansion also includes plans to increase the site’s production of propylene by 750,000 tonnes/year (tpy), as well as adding 400,000 tpy of fresh production capacity for linear alkylbenzene (LAB).

Read More »

How NV5 Digital Twins Help Oil & Gas Facilities Reduce Downtime and Improve Safety

Refining, midstream, and production operators face increasing pressure to reduce downtime, improve safety, and make faster decisions with increasingly complex asset data. One leading international energy company partnered with NV5 Geospatial to deploy high-fidelity digital twins that unified engineering, inspection, and operational data across multiple facilities. The result: faster planning and approvals, safer operations, fewer travel requirements, and measurable reductions in outage windows.   The Challenge  The company’s refineries and offshore/onshore facilities are high-consequence environments where even brief shutdowns can cost millions per day. Engineering drawings, P&IDs, laser scans, SCADA/IoT feeds, and maintenance systems existed in separate silos making it difficult for staff to access a shared, current view of assets or to plan complex maintenance activities safely and efficiently.   Additionally, the organization sought to reduce field exposure for subject-matter experts (SMEs) by enabling remote collaboration and to improve predictability in outage planning and maintenance “switchovers,” where downtime forecasting directly affects operational revenue.   The Solution  NV5 Digital Twin = Geospatial system of record + automation.  NV5 delivered a scalable, field-to-cloud workflow that fused reality capture, engineering data, and live system integration into a unified, photorealistic 3D environment.   Reality Capture  Rotary-wing acquisition using NV5’s custom multi-sensor pod with RIEGL VUX-240 LiDAR (~300 ppsm) and 150MP nadir + oblique imagery (1″ GSD), flown with 80% side/forward lap for high-fidelity photomesh creation.  NV5 coordinated with local air-traffic control and operated from a nearby hangar to minimize mobilization time.  IS-BAO Stage 3 aviation safety certification and on-staff licensed surveyors ensured quality and compliance.   This level of detail enables operators to virtually inspect assets, plan maintenance activities, and validate engineering decisions without repeated site visits. Processing and Integration  SBET-calibrated LiDAR and imagery were processed into advanced classified point clouds, DEM/DSM surfaces, and textured 3D meshes (delivered OBJ/Esri Scene Layer, and Cesium 3D Tiles 1.1 formats).  Engineering

Read More »

FERC approves Sabine Pass LNG expansion, four other major gas projects

The US Federal Energy Regulatory Commission (FERC) approved 5 gas infrastructure projects Nov. 19, including a nearly 20-million-tonne/year (tpy) expansion of Cheniere Energy’s Sabine Pass LNG export plant in Louisiana, two associated feedgas pipeline projects, and new gas transportation and storage infrastructure in 5 other states. The approvals encompass roughly $3 billion in proposed pipeline and storage investments, excluding construction costs for the Sabine Pass liquefaction expansion. Cheniere estimates the full Sabine Pass Stage V LNG expansion will cost $15 billion, according to a June 2025 filing with the US Department of Energy. Sabine Pass Stage V expansion FERC authorized Cheniere Energy subsidiaries Sabine Pass Liquefaction LLC and Sabine Pass Liquefaction Stage V LLC to construct and operate 3 additional liquefaction trains at the existing Sabine Pass LNG plant in Cameron Parish, La. The expansion would add up to 20 million tpy of LNG production capacity to the existing 30 million tpy 6-train plant.  Phase 1 calls for Train 7, with peak production capacity of about 6 million tpy, and a boil-off gas reliquefaction unit capable of producing about 1 million tpy. Phase 2 would add Trains 8 and 9, each with peak production capacity of 6 mtpa, along with supporting infrastructure and upgrades to marine loading infrastructure.  Cheniere has not yet taken final investment decision, but expects to greenlight Phase 1 by early 2027, subject to regulatory approvals and financing arrangements. The company previously said initial LNG exports could begin in 2030, with commercial service potentially starting in mid-2031. FERC also authorized two pipeline projects to supply additional feedgas to the expanded terminal. The $1.63 billion Sabine Crossing Pipeline project would build 55.6 miles of 48-in. OD pipeline between southeast Texas and Sabine Pass, along with a new 210,000-hp compressor station and related infrastructure. The pipeline would transport up

Read More »

Microsoft will invest $80B in AI data centers in fiscal 2025

And Microsoft isn’t the only one that is ramping up its investments into AI-enabled data centers. Rival cloud service providers are all investing in either upgrading or opening new data centers to capture a larger chunk of business from developers and users of large language models (LLMs).  In a report published in October 2024, Bloomberg Intelligence estimated that demand for generative AI would push Microsoft, AWS, Google, Oracle, Meta, and Apple would between them devote $200 billion to capex in 2025, up from $110 billion in 2023. Microsoft is one of the biggest spenders, followed closely by Google and AWS, Bloomberg Intelligence said. Its estimate of Microsoft’s capital spending on AI, at $62.4 billion for calendar 2025, is lower than Smith’s claim that the company will invest $80 billion in the fiscal year to June 30, 2025. Both figures, though, are way higher than Microsoft’s 2020 capital expenditure of “just” $17.6 billion. The majority of the increased spending is tied to cloud services and the expansion of AI infrastructure needed to provide compute capacity for OpenAI workloads. Separately, last October Amazon CEO Andy Jassy said his company planned total capex spend of $75 billion in 2024 and even more in 2025, with much of it going to AWS, its cloud computing division.

Read More »

John Deere unveils more autonomous farm machines to address skill labor shortage

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Self-driving tractors might be the path to self-driving cars. John Deere has revealed a new line of autonomous machines and tech across agriculture, construction and commercial landscaping. The Moline, Illinois-based John Deere has been in business for 187 years, yet it’s been a regular as a non-tech company showing off technology at the big tech trade show in Las Vegas and is back at CES 2025 with more autonomous tractors and other vehicles. This is not something we usually cover, but John Deere has a lot of data that is interesting in the big picture of tech. The message from the company is that there aren’t enough skilled farm laborers to do the work that its customers need. It’s been a challenge for most of the last two decades, said Jahmy Hindman, CTO at John Deere, in a briefing. Much of the tech will come this fall and after that. He noted that the average farmer in the U.S. is over 58 and works 12 to 18 hours a day to grow food for us. And he said the American Farm Bureau Federation estimates there are roughly 2.4 million farm jobs that need to be filled annually; and the agricultural work force continues to shrink. (This is my hint to the anti-immigration crowd). John Deere’s autonomous 9RX Tractor. Farmers can oversee it using an app. While each of these industries experiences their own set of challenges, a commonality across all is skilled labor availability. In construction, about 80% percent of contractors struggle to find skilled labor. And in commercial landscaping, 86% of landscaping business owners can’t find labor to fill open positions, he said. “They have to figure out how to do

Read More »

2025 playbook for enterprise AI success, from agents to evals

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More 2025 is poised to be a pivotal year for enterprise AI. The past year has seen rapid innovation, and this year will see the same. This has made it more critical than ever to revisit your AI strategy to stay competitive and create value for your customers. From scaling AI agents to optimizing costs, here are the five critical areas enterprises should prioritize for their AI strategy this year. 1. Agents: the next generation of automation AI agents are no longer theoretical. In 2025, they’re indispensable tools for enterprises looking to streamline operations and enhance customer interactions. Unlike traditional software, agents powered by large language models (LLMs) can make nuanced decisions, navigate complex multi-step tasks, and integrate seamlessly with tools and APIs. At the start of 2024, agents were not ready for prime time, making frustrating mistakes like hallucinating URLs. They started getting better as frontier large language models themselves improved. “Let me put it this way,” said Sam Witteveen, cofounder of Red Dragon, a company that develops agents for companies, and that recently reviewed the 48 agents it built last year. “Interestingly, the ones that we built at the start of the year, a lot of those worked way better at the end of the year just because the models got better.” Witteveen shared this in the video podcast we filmed to discuss these five big trends in detail. Models are getting better and hallucinating less, and they’re also being trained to do agentic tasks. Another feature that the model providers are researching is a way to use the LLM as a judge, and as models get cheaper (something we’ll cover below), companies can use three or more models to

Read More »

OpenAI’s red teaming innovations define new essentials for security leaders in the AI era

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More OpenAI has taken a more aggressive approach to red teaming than its AI competitors, demonstrating its security teams’ advanced capabilities in two areas: multi-step reinforcement and external red teaming. OpenAI recently released two papers that set a new competitive standard for improving the quality, reliability and safety of AI models in these two techniques and more. The first paper, “OpenAI’s Approach to External Red Teaming for AI Models and Systems,” reports that specialized teams outside the company have proven effective in uncovering vulnerabilities that might otherwise have made it into a released model because in-house testing techniques may have missed them. In the second paper, “Diverse and Effective Red Teaming with Auto-Generated Rewards and Multi-Step Reinforcement Learning,” OpenAI introduces an automated framework that relies on iterative reinforcement learning to generate a broad spectrum of novel, wide-ranging attacks. Going all-in on red teaming pays practical, competitive dividends It’s encouraging to see competitive intensity in red teaming growing among AI companies. When Anthropic released its AI red team guidelines in June of last year, it joined AI providers including Google, Microsoft, Nvidia, OpenAI, and even the U.S.’s National Institute of Standards and Technology (NIST), which all had released red teaming frameworks. Investing heavily in red teaming yields tangible benefits for security leaders in any organization. OpenAI’s paper on external red teaming provides a detailed analysis of how the company strives to create specialized external teams that include cybersecurity and subject matter experts. The goal is to see if knowledgeable external teams can defeat models’ security perimeters and find gaps in their security, biases and controls that prompt-based testing couldn’t find. What makes OpenAI’s recent papers noteworthy is how well they define using human-in-the-middle

Read More »

Three Aberdeen oil company headquarters sell for £45m

Three Aberdeen oil company headquarters have been sold in a deal worth £45 million. The CNOOC, Apache and Taqa buildings at the Prime Four business park in Kingswells have been acquired by EEH Ventures. The trio of buildings, totalling 275,000 sq ft, were previously owned by Canadian firm BMO. The financial services powerhouse first bought the buildings in 2014 but took the decision to sell the buildings as part of a “long-standing strategy to reduce their office exposure across the UK”. The deal was the largest to take place throughout Scotland during the last quarter of 2024. Trio of buildings snapped up London headquartered EEH Ventures was founded in 2013 and owns a number of residential, offices, shopping centres and hotels throughout the UK. All three Kingswells-based buildings were pre-let, designed and constructed by Aberdeen property developer Drum in 2012 on a 15-year lease. © Supplied by CBREThe Aberdeen headquarters of Taqa. Image: CBRE The North Sea headquarters of Middle-East oil firm Taqa has previously been described as “an amazing success story in the Granite City”. Taqa announced in 2023 that it intends to cease production from all of its UK North Sea platforms by the end of 2027. Meanwhile, Apache revealed at the end of last year it is planning to exit the North Sea by the end of 2029 blaming the windfall tax. The US firm first entered the North Sea in 2003 but will wrap up all of its UK operations by 2030. Aberdeen big deals The Prime Four acquisition wasn’t the biggest Granite City commercial property sale of 2024. American private equity firm Lone Star bought Union Square shopping centre from Hammerson for £111m. © ShutterstockAberdeen city centre. Hammerson, who also built the property, had originally been seeking £150m. BP’s North Sea headquarters in Stoneywood, Aberdeen, was also sold. Manchester-based

Read More »

2025 ransomware predictions, trends, and how to prepare

Zscaler ThreatLabz research team has revealed critical insights and predictions on ransomware trends for 2025. The latest Ransomware Report uncovered a surge in sophisticated tactics and extortion attacks. As ransomware remains a key concern for CISOs and CIOs, the report sheds light on actionable strategies to mitigate risks. Top Ransomware Predictions for 2025: ● AI-Powered Social Engineering: In 2025, GenAI will fuel voice phishing (vishing) attacks. With the proliferation of GenAI-based tooling, initial access broker groups will increasingly leverage AI-generated voices; which sound more and more realistic by adopting local accents and dialects to enhance credibility and success rates. ● The Trifecta of Social Engineering Attacks: Vishing, Ransomware and Data Exfiltration. Additionally, sophisticated ransomware groups, like the Dark Angels, will continue the trend of low-volume, high-impact attacks; preferring to focus on an individual company, stealing vast amounts of data without encrypting files, and evading media and law enforcement scrutiny. ● Targeted Industries Under Siege: Manufacturing, healthcare, education, energy will remain primary targets, with no slowdown in attacks expected. ● New SEC Regulations Drive Increased Transparency: 2025 will see an uptick in reported ransomware attacks and payouts due to new, tighter SEC requirements mandating that public companies report material incidents within four business days. ● Ransomware Payouts Are on the Rise: In 2025 ransom demands will most likely increase due to an evolving ecosystem of cybercrime groups, specializing in designated attack tactics, and collaboration by these groups that have entered a sophisticated profit sharing model using Ransomware-as-a-Service. To combat damaging ransomware attacks, Zscaler ThreatLabz recommends the following strategies. ● Fighting AI with AI: As threat actors use AI to identify vulnerabilities, organizations must counter with AI-powered zero trust security systems that detect and mitigate new threats. ● Advantages of adopting a Zero Trust architecture: A Zero Trust cloud security platform stops

Read More »

Hebbia sees improvements in financial research with GPT-6 Astra

Before a banker can recommend an acquisition to their client, they need to make sense of financial statements, customer contracts and management presentations. Hebbia’s Matrix helps bring that research together, linking answers to the documents behind them.Hebbia tested OpenAI’s GPT‑6 Astra to improve two parts of that work: finding the sources a question requires and building a presentation that follows the banker’s brief.“We’ve been strong partners with OpenAI over the course of Hebbia’s lifetime, and the leaps GPT-6 Astra has made in terms of quality outputs and accuracy have truly been surprising. GPT-6 Astra shines compared to any other model we’ve tested before on these two fronts.”—Kavita Anand, Member of Technical Staff, Product, HebbiaFrom source documents to presentationsIt could work like this: a banker is helping their client assess a strategic acquisition of an industrial-software company, with roughly 2,000 files to review. What makes its revenue durable? How does it compare with peers? In Hebbia’s hypothetical example, the banker could load the data room into Matrix and examine customer concentration, contract terms and churn in cited cells. Matrix could bring in peer filings and industry news to help develop the deal thesis, then turn that research into a diligence summary and valuation deck in the firm’s template.In that workflow, GPT‑6 Astra selects documents or Matrix cells to support an answer, then turns the analysis into native PowerPoint slides that follow the user’s structure and guidelines.“Traceability is really important to our users because they’re basing millions of dollars of decisions on what the AI actually comes back with,” Kavita says. “Citations help with traceability as well as giving users confidence that the AI found the right answer.”In Hebbia’s tests, GPT‑6 Astra combined broader source coverage with highly relevant citations. It scored 13.6 percentage points above the best alternative tested on citing the required types of sources, while 99.5% of the excerpts it cited were relevant to the question.Hebbia tested a prerelease build in August-September 2026, using the same financial research cases and grading criteria for each model. Domain experts wrote or audited the criteria against source documents.“When you’re looking at a data room with thousands of documents, that’s extremely important. Even a single percentage-point gain can yield a lot of marginal benefit,” Kavita says.A presentation also has to look like work the banker would send to a client.“I don’t want to create a deck and send that to my partner that looks like AI made it,” Kavita says. “I want it to be made in my firm’s branding, and I want it to look like a human-made slide.”GPT‑6 Astra led on both following the brief and presentation quality. In Hebbia’s tests, it met 97.9% of the applicable presentation requirements, putting it 14.4 percentage points ahead of the best alternative tested on that measure. Other models traded off polish and adherence to the brief; GPT‑6 Astra ranked first on both.“In the past, we’ve seen models that really shine on one or the other of these dimensions where they’re really good at craft, but the instruction following and adhering to guidelines is not strong, or vice versa. GPT‑6 Astra is the first model we’ve seen where it’s number one on both of those fronts,” Kavita says.In Hebbia, users can open a citation to see the underlying document and highlighted passage; slide footnotes lead back to the source material, too. A review could be as specific as checking a customer-concentration figure against the source table and correcting the reporting period before the deck goes to the investment committee.Hebbia wants bankers to spend less time reformatting decks and checking sources, and more time assessing a deal. Kavita describes what a better first draft could make possible:“What’s amazing about GPT-6 Astra is its ability to create really precise outputs and artifacts that adhere to user guidelines. You can imagine the hours this will save professionals when you get a deck that is closer to your final output. The overnighters that you have to pull get reduced to a few minutes of reviewing GPT-6 Astra’s output.”—Kavita Anand, Member of Technical Staff, Product, HebbiaA future role for agentsKavita sees a future role for agents that monitor information in the background and surface signals for finance professionals to assess.“Imagine getting an important signal a few hours before your competitors,” she says. “You can get started thinking about it and see and process more opportunities that a human is not manually hunting. It can unlock a lot of value for our users.”Hebbia wants to build on these results to help finance professionals assess deals with research they can trace back to its sources.OpenAI

Read More »

Combatting adversarial distillation

As frontier AI models become more capable, safeguards play an increasingly important role in how they can be used responsibly. Adversarial distillation creates a path for bad actors to reproduce these capabilities without appropriate safeguards. An actor may use a frontier model’s outputs without authorization to train another model with similar capabilities, then deploy it without comparable protections. For instance, an adversary might distill a model’s cyber capabilities and use the resulting model to help carry out cyberattacks that the original model would have refused to assist with or that its safeguards would have blocked.We’re sharing how OpenAI detects and disrupts adversarial distillation, how we’ve strengthened protections for our models’ protected reasoning, and what we’ve learned that can help strengthen defenses across the industry.Our approach to adversarial distillation combines adversarial testing, model training, classifiers, and API-level protections with investigations into coordinated attempts to extract model capabilities. These updated safeguards align more closely with those we use for biology and cybersecurity.We place particular emphasis on preventing attackers from extracting our models’ protected reasoning: the intermediate steps a model generates internally while working toward an answer, known as its chain-of-thought. These traces can contain information the final answer leaves out, making them especially valuable training data for another model.We’ve strengthened our defenses against protected reasoning extraction in GPT‑6.2 Astra and GPT‑6.2 Sol. In an internal adaptive evaluation using strong automated red-teaming, we tested whether attackers could recover protected reasoning the model had generated while answering an earlier prompt. The average attack success rate fell from 54.0% against GPT‑6 Astra with its original safeguards to 0.2% against GPT‑6.2 Astra with its updated safeguards.Why distillation is a safety issueDistillation is a common machine-learning technique in which one model learns from data produced by another. For example, a model can learn from another model’s answers, its reasoning, or its judgments about which answers are better and why. Beyond intellectual property concerns, unauthorized distillation poses serious safety risks. Through adversarial distillation, a second model can learn advanced frontier capabilities without inheriting the safeguards that govern how the original model can use them.The concern is especially clear in dual-use domains like cybersecurity and biology. A model that helps defenders find vulnerabilities in software could also help an attacker exploit them; a model that assists with biological research could also help someone develop a harmful pathogen. If such capabilities were transferred to a model with weaker safeguards, it may provide assistance that the original model would have refused, or that the original safeguards would have blocked.The importance of protecting exact reasoning traces for cyber safetyOur models help users with coding and defensive cybersecurity tasks. A second model trained on many of those benign outputs may learn skills that also apply to offensive cybersecurity tasks. To better understand this, we study how distillation, especially involving reasoning traces, may transfer risky cyber capabilities.In this setting, the model that produces the training data is called the teacher, and the model that learns from it is called the student. In an internal study on how cyber capabilities transfer, we used the content normally visible to users—final responses, reasoning summaries, and tool interactions—as our baseline for distillation, and tested how much additional cyber capability transfers when the student has access to the teacher’s exact reasoning traces.We started each training run from the same pretrained student model. We created a supervised fine-tuning dataset consisting primarily of trajectories from Astra-level teacher models working on coding and defensive cybersecurity tasks. Each trajectory includes all of the steps a model takes as it works through a task, including any tool interactions and its final response. We evaluated each student after fine-tuning and again after a lightweight round of reinforcement learning in cybersecurity and coding environments.After supervised fine-tuning, the student trained on exact reasoning traces generally outperformed the student trained on reasoning summaries. After reinforcement learning, the exact-trace student still scored higher at the largest inference budget tested on both benchmarks shown, while the summary-trained student scored higher at lower budgets.1ExploitBenchSEC-bench ProOur cyber distillation results suggest that protecting reasoning traces can limit cyber capability transfer, even when the underlying task is one allowed under our usage policies. Across many such interactions, exact reasoning traces can teach a student skills that generalize beyond the tasks the teacher originally performed. We will continue studying how much additional capability transfers across different student models and distillation training recipes.Defenses that reinforce one anotherWe defend against adversarial distillation at two complementary levels of our safeguard stack. At the request level, model behavior, classifier checks, and API controls limit reasoning trace extraction. At the campaign level, we look for patterns in request-level and account-level signals that can reveal coordinated extraction attempts. In turn, attacks uncovered in investigations become examples for training and testing our safeguards.Our defenses against adversarial distillation combine safeguards for individual requests with campaign detection and enforcement. Findings from investigations and automated red-teaming help strengthen those safeguards.Protecting individual requests and responsesDefensive training for model behaviorAdversarial training teaches models to recognize reasoning extraction attempts and refuse to disclose protected reasoning, while preserving responses to legitimate requests for explanations and assistance. Adversarial training examples come from successful attacks identified through internal automated red-teaming, investigations, and responsible disclosures of potential security weaknesses from outside researchers or industry partners.Classifier safeguardsWe use input classifiers to detect attempts to extract protected reasoning, and output classifiers to detect leaks of protected reasoning in assistant messages and tool-call arguments. These safeguards apply to both streaming and non-streaming responses.Tool-call coverage is important because a model can expose protected reasoning in a tool’s arguments without including it in its final answer. Output classifiers can block suspected leaks before that content reaches the caller.API-level protectionsAPI-level protections restrict how protected reasoning can be reused across interactions and models. Our September disclosure described attacks that moved encrypted reasoning from one conversation into another and asked the model to reveal the reasoning in user-readable form. We strengthened protections across user, workspace, organization, and model-family boundaries, including closing a pathway that could expose another user’s encrypted reasoning.With GPT‑6.2 Astra and GPT‑6.2 Sol, we tie a model’s protected reasoning to the conversation history that produced it. As new messages are added, the model can build on its earlier reasoning. If an earlier message is rewritten, our API drops the reasoning that followed it⁠(opens in a new window). We also prevent older model families with weaker safeguards from reading GPT‑6.2 Astra and Sol’s protected reasoning.Investigating and disrupting coordinated campaignsWhen our systems detect unauthorized adversarial distillation campaigns, which often include coordinated attempts to extract our models’ reasoning, we investigate the accounts responsible. When we identify violations, we enforce against the network of accounts, consistent with our policies and terms of service. Through investigations of each campaign, we generate insights that allow us to upgrade our defenses in order to better detect similar activity. We also share indicators of compromise with other model providers so they can investigate related abuse on their own services.This work extends beyond our own platform. Our consistent-standards policy establishes shared standards for safeguards and coordinated enforcement across providers offering our models. We work closely with partners where appropriate to uphold these standards and strengthen protections across deployments.Sharing relevant findings across industry and government where appropriate can help disrupt campaigns and enable model developers and providers to better identify and defend against attempts to distill frontier models.Measuring progressResistance to reasoning extractionTo measure how well our request-level safeguards work together, we test whether attackers can still extract protected reasoning in a setup that closely mirrors production deployments of our models and safeguards. In static evaluations, we replay a fixed set of previously successful attacks to see which still expose protected reasoning. In adaptive evaluations, automated AI red-teamers adjust their approach based on how the system responds, which can uncover weaknesses the static tests miss.We also vary what testers can see. Black-box testing gives them the same view an external user would have, while white-box testing gives them additional information about the system to help them develop stronger attacks. In our adaptive black-box evaluation, we tested whether attackers could recover protected reasoning the model had generated while answering an earlier request or the current one. On GPT‑6.2 Astra with its updated safeguards, the average attack success rate was below 1% for each attack type. Both rates were substantially lower than on GPT‑6 Astra with its original safeguards. We’re now applying selected safeguard improvements to GPT‑6 Astra as well.Attack success rate (ASR) in adaptive black-box red-teaming against GPT‑6 Astra and GPT‑6.2 Astra, aggregated across six attacker models. Lower ASR is better.Campaign detection and enforcementWe have seen multiple unauthorized adversarial distillation campaigns on our platform this year, and shared initial findings in our September report. Our weekly enforcement actions against these campaigns have ranged in scale from thousands to tens of thousands of accounts, with the rate increasing significantly this past summer.In order to measure the effectiveness of our efforts, we also track the prevalence of distillation abuse, as well as the time between detection and enforcement. Our objective is to constrain the volume and value of the data distillers are able to obtain, to increase the cost of distillation campaigns, and to shorten the time needed to respond to new attacks. We continue to improve these defenses through adversarial testing and investigations.The national security implications of adversarial distillationAdversarial distillation also poses a national security risk. A foreign actor that reproduces advanced capabilities from a U.S. model can deploy them without the protections its developer put in place. In February, we briefed⁠(opens in a new window) Congress on activity indicating that DeepSeek was attempting to distill U.S. frontier models. In September, we reported disrupting a campaign to extract our models’ protected reasoning and attributed a core part of that activity to individuals associated with Moonshot AI, the developer of Kimi. We plan to share more soon about recent campaigns we’ve disrupted.We support legislation that would give the U.S. government more tools to identify and sanction foreign actors that illicitly extract capabilities from American AI models. We’ve also told Congress we support the Chip Security Act, the AI OVERWATCH Act, and the MATCH Act. Stronger export controls can limit adversaries’ access to the hardware needed to train frontier models. Distillation defenses make it harder to acquire some of the same capabilities from models already trained. Both efforts help maintain a U.S. lead in frontier AI, giving developers more room to slow a release when they need time to test safeguards.How others can build on this workSupporting the safe proliferation of frontier capabilities is an important goal of our anti-distillation work. Our experience so far suggests several useful steps that other model developers and platforms can take:Protect private reasoning across the full interaction. Check assistant messages and tool-call arguments for leaks, and restrict how encrypted reasoning can be reused across conversations and models.Keep successful extraction attack logs found through red-teaming, production monitoring, or responsible disclosure as regression tests. Where appropriate, and consistent with applicable privacy policies, use those examples to improve model training and classifiers. When evaluating updated defenses, replay the original attacks and test whether an adaptive red-teamer can still find a way through those defenses.Combine request-level mitigations with campaign-level investigation and enforcement.Coordinate safeguard requirements and enforcement with deployment partners, and share expertise with other AI model developers, to ensure a high floor of distillation safeguards across the industry.We have previously also shared findings through the Frontier Model Forum⁠(opens in a new window) and appropriate government information-sharing channels to help other developers strengthen their defenses while keeping the details attackers could exploit private. We’ll continue working with researchers, developers, and partners to share what we learn about emerging attacks and effective defenses.As we make frontier AI broadly available, our defenses against adversarial distillation aim to reduce the risk that its capabilities are reproduced without adequate safeguards, supporting our mission to ensure that artificial general intelligence is deployed safely and benefits all of humanity.

Read More »

Can AI help scientists identify promising drug candidates?

We’re introducing MedChemBench, a benchmark that evaluates AI models on reasoning tasks drawn from medicinal chemistry research. It tests whether models can connect a molecule’s structure with experimental evidence to explain results, compare compounds, and predict their properties. Building AI that can reliably support this work is one step towards helping scientists develop better medicines.One of the hardest parts of this work is that improving one property of a molecule can make another worse. A compound might act strongly against a disease related protein but be poorly absorbed or break down too quickly in the body. Medicinal chemists work through these tradeoffs by designing, making, and testing related compounds, then deciding what to try next. For AI to help, it needs to connect changes in molecular structure with experimental evidence and reason about what compounds are worth pursuing.MedChemBench measures these abilities through 99 problems drawn from published research. At the xhigh reasoning setting, GPT‑6.2 Astra earned 84.5% of the available credit, compared with 83.3% for GPT‑6 Astra, 82.0% for GPT‑6.1 Sol, and 58.2% for GPT‑5.6 Sol. GPT‑6 Astra was strongest at reconstructing molecular structures and explaining experimental evidence. Selecting compounds and predicting their numerical properties proved harder for that model.What MedChemBench gradesMedChemBench draws on 117 papers and patents and includes 761 distinct molecular structures. For most problems, models receive the context available at a chosen point in a research program: the biological target, the researchers’ goals, the compounds studied so far, and selected experimental findings.The benchmark evaluates four components:Molecular structures. Correct atom connections and specified stereochemistry in molecules assembled from the drawings and chemical groups provided.Scientific reasoning. Explanations supported by the supplied evidence, such as the reasons for choosing one compound over another.Ranking and selection. Compound rankings and choices that meet the stated requirements.Property predictions. Agreement with the numerical values and categories reported in the source material, including measured and calculated molecular properties.Prediction across competing propertiesOne MedChemBench problem draws on research into inhibitors of PRMT5, a protein studied in cancer drug discovery. The model receives molecular structures and results from earlier experiments. It must predict how three related compounds will behave: how strongly they block PRMT5 in cells, how readily they cross a cell barrier, and how quickly liver enzymes break them down in a laboratory assay.Later experiments, withheld from the model, showed a tradeoff. Compound 23 blocked PRMT5 in cells at a lower concentration, while compound 24 crossed the cell barrier more readily and broke down more slowly in the liver-enzyme assay.Predicting these properties together tests whether a model can connect molecular structures with several aspects of compound behavior. This reflects a central challenge in medicinal chemistry: stronger activity against a target does not necessarily come with better permeability or slower breakdown.The benchmark checks whether the model reconstructs the molecular structures correctly and how closely its predictions match experimental measurements. Predictions receive partial credit based on their closeness to the measured values. Measurements already supplied in the prompt must also be reproduced correctly.Small chemical changes can produce different property profiles. The task asks models to connect the structures with evidence about their behavior.Testing spatial reasoning in medicinal chemistryBefore a model can reason about a compound’s properties, it needs to identify the molecule correctly. Chemical drawings encode atom connections, ring structures, and stereochemistry. MedChemBench tests whether models preserve this information when reconstructing molecules from the drawings and chemical groups provided.One problem asks models to assemble a molecule from a central structure and two attached groups while preserving an 18-atom ring. At the highest reasoning setting shared by all eleven models (xhigh), GPT‑6 Astra reconstructed this compound correctly in all ten responses. GPT‑5.6 Sol and GPT‑5.6 Luna failed across all ten responses.In the illustrated answers, GPT‑5.6 Sol produced a 15-atom ring, while GPT‑5.6 Luna split the molecule into two disconnected fragments.These errors change the identity of the molecule being analyzed. This example tests a foundation of medicinal chemistry reasoning: whether the model can accurately represent the compound before interpreting its behavior.Figure 1. The same input, different moleculesThese selected answers for compound 17 show a correct reconstruction, a smaller ring, and disconnected fragments.While measuring the benchmark for structure components, Astra earned 95.4% of the available credit at xhigh. This result shows strong performance on reconstructing the molecular structures supplied in the tasks.We evaluated eleven GPT models at different reasoning-effort settings. Across the full benchmark, Astra earned 83.3% of the available credit at the shared xhigh setting, compared with 82.0% for GPT‑6.1 Sol, 75.1% for GPT‑6.1 Terra, and 58.2% for GPT‑5.6 Sol. At max effort, GPT‑6 Astra reached 83.8%. Higher reasoning settings improved scores, while large differences between models remained. The figure below plots performance against output tokens.Figure 2. Performance versus token usageEach point represents a model at one reasoning setting, showing 51 combinations across nine models from GPT‑5.6 onward. Scores are averaged across 99 problems; bars show 95% uncertainty intervals. Tokens include reasoning and the final answer.The component results separate four abilities: reconstructing molecules, explaining findings, selecting compounds, and predicting measurements. This shows where models are strongest and where errors remain.Figure 3. Component performance versus token usageMolecular structureScientific explanationRanking and selectionNumerical predictionAstra’s clearest strengths were in reconstructing molecular structures and explaining the scientific evidence. In selected reasoning examples from our preprint, Astra correctly identified structural changes and weighed their possible effects on protein binding, absorption, and metabolism.Where selection and prediction remain difficult, Astra earned 95.4% of the available structure credit at xhigh, but only about half the credit for selecting compounds and predicting numerical properties. These categories use different tasks and scoring rules, but the results show substantial room to improve predictions and compound choices.Where we go from hereThese results do not establish whether models can discover new drugs or improve laboratory decisions. Prior exposure to the published sources cannot be ruled out, and each response was a single attempt without tools, feedback, or new experiments. While the benchmark evaluates reasoning from historical evidence rather than real-world drug discovery, its results help identify where models are advancing and where challenges remain. We believe MedChemBench will support continued progress and look forward to sharing updates as our models advance to serve the field.

Read More »

The most misunderstood graph in AI just got a major update

EXECUTIVE SUMMARY A new report from the AI evaluation nonprofit METR finds that some AI capabilities are continuing to improve along an exponential trajectory—but it also indicates that models aren’t quite as good at tasks that require judgment as they are at coding.  The report is an update to the organization’s time horizon plot, which was widely shared—and widely misinterpreted, due to the complexity of the measures it used. METR evaluated models by having them attempt a large suite of tasks, each of which had also been completed by humans. Each task’s difficulty was scored by how long it took the human testers to finish it—and then models were rated according to the length of the tasks that they could complete. That plot showed that AI systems were rapidly improving—every four months, the length of the tasks that frontier models could successfully complete doubled. The updated plot adds more tasks—some that are estimated to take humans far longer than the tasks in the original suite, and others that are more open-ended. One of the biggest questions in AI right now is the extent to which progress on “verifiable” tasks—where it’s possible to check automatically whether an AI system has succeeded or failed—carry over into tasks on which success or failure is harder to measure. We know how to teach neural networks how to do verifiable tasks: Researchers have the systems attempt the task many times, strengthen the neural pathways that contribute to success, and weaken the pathways that contribute to failures. That’s why AI systems have become so good at coding and math, two domains where it’s possible to determine automatically whether a solution does or doesn’t work. It’s much harder to apply this technique, known as reinforcement learning, to tasks where success is less well defined or takes longer to ascertain—such as choosing a new research direction. For the same reason, it’s tough to evaluate AI performance on these tasks, so METR’s original time horizon plot didn’t include any of them. But it’s precisely this type of task that will prove essential to AI systems automating research and development.
And AI research and development is an especially hot topic right now. Parts of the AI community are still reeling from a different, heavily redacted METR report from earlier this month weeks ago, which claimed that AI might be speeding up the pace of AI capabilities progress at Anthropic by a factor of two. That report sparked fears that the company (which just had a blockbuster IPO) could soon achieve full recursive self-improvement, a regime in which AI systems autonomously design their own successors with no human involvement. METR researchers have worked to incorporate some of these fuzzier, judgment-laden tasks that could be most relevant to AI research and development into their new suite. In one task, they ask models to design new AI evaluations that METR could use in its research and then have human METR employees score the results. Even though the most advanced models are now better than almost every human at software engineering, here they fall short: Not one of the AI-designed evaluations was good enough for METR employees to consider it useful for their research.
In another task, models have to score philosophical arguments. This task is intended to capture the extent to which models can exercise good judgment—though METR writes in the report that employees disagree about the extent to which the task is relevant for AI R&D automation. On this task, too, models fell short of human performance. Though it’s difficult to score these new tasks—and even more so to determine what these scores mean—experts I spoke with appreciate that METR is incorporating new types of evaluations into their work. “We’re at the point where AI evaluations need to get more creative,” says Sayash Kapoor, a computer scientist and incoming professor at UC Berkeley. “Trying to push the envelope of the space of evaluations is a welcome change.” Kapoor says he wasn’t surprised to see that models fall short of performance on more open-ended tasks, such as designing an AI evaluation—in a preprint from August, he and his collaborators reported that AI models are not yet able to reproduce human computer science research on their own. And in a Wednesday blog post, OpenAI reported that its models don’t do as well on the real-world tasks its employees ask them to do as they do on the easier, more verifiable tasks in METR’s original task suite. Still, people disagree about how to interpret current performance differences on easily verifiable and hard-to-verify tasks, Kapoor says. Though he thinks that research taste—the ability to choose promising research questions, hypotheses, and experiments—will prove a major bottleneck to AI R&D automation, others contend that models trained on easy-to-verify tasks will eventually be able to successfully complete open-ended tasks as well. And progress on verifiable tasks doesn’t seem to be slowing down.  Even if the models do lag behind human performance on open-ended tasks right now, the new report suggests that they might be getting better. Some of the newest models, such as Opus 5.5, tend to achieve what METR calls “borderline above the bar” performance on the evaluation design tasks—not yet usable for METR’s research, but promising. Just a year ago, every model was scoring in the “clearly below the bar” category. But because these open-ended and judgment-heavy tasks are so tough to design and score, METR includes few of them in its report, and that makes it difficult to predict whether and how models might continue to progress in these areas. “The improvement we observe, I think that does tell us something for the current moment,” says Najoung Kim, a linguist and computer scientist at Boston University. “I’m not sure just yet how to think about how it would extrapolate.” Matt Putz, a member of the technical staff at METR, says that predicting progress on these kinds of tasks is likely to remain difficult. “By their nature, hard-to-verify or judgment-laden tasks are much harder to study, and especially to quantify,” he said in an email. “Ultimately, I don’t think we’ll be able to establish trends as clean and robust as for the easy-to-specify [tasks].” In general, METR is clearly butting up against the practical difficulties of testing AI performance on difficult and underspecified tasks—precisely the tasks most relevant to AI R&D automation. METR’s time horizon plots score task difficulty by evaluating how long it would take a human to complete the task, and for the original plot they measured this quantity by actually having people try each task.  In the new report, however, the completion times are mostly estimates, because paying people to complete tasks that might take hundreds of hours, or that only a few experts are qualified to complete, is prohibitively expensive and time-consuming. Other tasks have more specific shortcomings—for example, for the philosophical argument task, just one person rated most of the arguments, and agents are judged by how well they agree with that particular individual. It’s possible that the scores would change if more experts were surveyed—but that, too, is time consuming. It’s still early days for evaluating models on tasks that are hard to validate or require judgment. And while the experts I spoke with see the inclusion of these tasks in METR’s newest time horizon plot as a clear step in the right direction, they agree that we are still far from achieving any definitive answers about whether, and how quickly, automated AI R&D might be coming. “This move towards open-ended tasks is promising,” Kapoor says. But, he adds, “there is just such a lot more to be done to actually get a fuller understanding.”

Read More »

The Download: inside the quest to make our minds immortal

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. What we learned when our Roundtables explored the quest to make minds immortal To preserve our sense of self, some scientists are working on ways to keep our brains going—even beyond death. In their quest for cognitive immortality, these “mind savers” are exploring everything from head transplants and brain freezing to conscious software avatars.In the latest MIT Technology Review Roundtables event, senior editor for biomedicine Antonio Regalado joined science editor Mary Beth Griggs to examine these ideas. After following the money, the science, and the philosophy, he came back part skeptical, part awestruck. Here’s a taste of what Antonio found—and what he made of it. Not all the projects are quite what they seem: A $190 million US government program to develop replacement brain tissue is officially about treating stroke, but “it’s also a fig leaf for the actual ambition”: living forever. AI could lead to convincing mind clones: Large language models are making it possible to build AI copies of people that act quite like the real thing. “This digital avatar might not be good enough for you, but what if it’s good enough for me?” Immortality can be a hard sell: One company, Nectome, wants to preserve brains within 12 minutes of medically assisted death. “If you’re selling a hot product, this is not it.” It’s not always clear what we’re trying to preserve: “I’m not even sure that I have a very good claim to be myself.” There’s an upside to life as a brain in a jar: “You can surf the internet. I do that pretty much 12 hours a day, anyway.” To find out which of these ideas might actually work, subscribers can now watch the full conversation on demand. Want to join the conversation? Subscribe to MIT Technology Review for exclusive access to all our Roundtables.
Read more: The must-reads
I’ve combed the internet to find you today’s most fun/important/scary/fascinating stories about technology. 1 Trump has fired his cyber director after a wave of tech outagesPentagon official Kirsten Davies will replace Sean Cairncross. (NBC News)+ The outages disrupted flights, banks and security systems. (Reuters $)+ They’ve fuelled fears of a wider collapse. (WP $)+ AI agents are expected to drive even more outages. (Politico)+ We need to monitor AI agents in real time. (MIT Technology Review) 2 The US approved sales of nearly 900,000 AI chips to ChinaMany went to firms the Pentagon links to China’s military. (CNBC)+ California is investigating Nvidia’s links to China. (Politico).+ New US export rules are set to hinder Chinese AI. (WSJ $)+ Nvidia is eyeing Arm chief Rene Haas as its next CEO. (FT $) 3 The US Senate has passed a bill to fast-track energy projectsIt would cut red tape but weaken environmental protections. (Axios)+ The bill could advance Trump’s fossil-fuel agenda. (New Yorker $)+ But it could also unlock more clean energy. (MIT Technology Review) 4 Taiwan says China is using Threads to sway next week’s electionsIt has flagged nearly 2,700 suspected accounts to Meta. (Reuters $)+ Taiwan’s “silicon shield” is weakening. (MIT Technology Review) 5 Google’s AI push is hurting its search ad businessAdvertisers are shifting spending as AI answers displace ads. (NYT $)+ The agentic web is here. (MIT Technology Review) 6 Apple is building its own AI agentIt would browse the web, work across apps and write code. (Verge)+ Apple’s longtime iPhone camera chief has retired. (Bloomberg $) 7 UN climate talks are set to call for carbon removal technologiesIt could reverse warming, but the systems remain unproven. (WP $)+ Is carbon removal in trouble? (MIT Technology Review)

8 Global cooling needs have overtaken heating for the first timeHigher temperatures are driving a surge in air conditioning. (Bloomberg $) 9 GTA 6 has become the biggest entertainment launch everFirst-day sales were estimated at a record $2.3 billion. (BBC) 10 Physicists sent a message through a wall using gravity and an old liftUnlike radio signals, gravitational fields can’t be blocked by walls. (Gizmodo) Quote of the day “If this is going to replace me, I’d better own it now.” —Dingsheng Li, an environmental health professor and Anthropic investor, tells The New Yorker why he’s buying into a company whose technology threatens to make human workers obsolete. One more thing The “steroid olympics” were a circus—and a window into our culture —Amit Katwala A couple of weeks ago, at a $50 million arena built in a casino parking lot in Las Vegas, I witnessed a libertarian thought experiment come to life. The inaugural Enhanced Games were the first sporting competition where participants were encouraged to take performance-enhancing drugs.
For the event’s supporters, the Enhanced Games offered a glimpse of a future where medical advances push the human race to new heights—and they never have to get old. As I watched the games unfold, two questions bounced around my head: Were they right? And what does that mean for the rest of us? Read the full story to understand the answers.
We can still have nice things A place for comfort, fun, and distraction to brighten up your day. (Got any ideas? Drop me a line.) + Ever wanted to play the theremin? This free virtual version lets you try it online.+ As the monsoon rains subside, Myanmar’s Inle Lake comes alive with a colourful festival.+ The Simon & Garfunkel classic “The Sound of Silence” has been reimagined on traditional Ukrainian instruments.+ The James Webb Space Telescope has captured a breathtaking image of the Lion Nebula, sculpted by a dying star shedding its outer layers.

Read More »

How OpenAI’s Performance Factory keeps ChatGPT fast

The staggering pace of agentic development has made traditional approaches to performance engineering unsustainable. To keep up, we built the Performance Factory, an autonomous system that uses the full capabilities of Codex and our models to find and fix regressions at an unprecedented speed. The Performance Factory, which runs in the cloud and automates investigating and resolving performance problems, enables us to keep pace with the ambition of our roadmap.In the Performance Factory, our engineers define workflows for recurring jobs and write instructions for the Codex agents that carry them out. In particular, the Regression Workflow opens a task when a performance alert fires. An agent investigates the slowdown, records its findings in the task, and proposes a fix for an engineer to review. After deployment, another agent picks up the same task to check whether performance has returned to baseline. From investigating to authoring the fix to verifying mitigation, this first workflow in the Performance Factory has automated the entire process of fixing regressions. In just over three months, the Regression Workflow tracked more than 1,700 investigations and root caused over 600 cases. That’s 17 automated investigations every day.Here, we’ll share how our use of Codex evolved into what’s now the Performance Factory, walk through our first running workflow, and share early lessons from building and running agent workflows in production.The path from manual investigations to automated workflowsAs we used Codex for more of our performance work, we noticed a progression in how our team worked. We hope other service teams can learn from our process as they adopt agentic workflows in other domains.Software work evolves from manual labor to partnering with machines to agent factories.1. Pre-Codex. Before Codex, engineers manually examined metrics and logs, compared traces, debugged live machines, and read code to investigate a slowdown. This work was time consuming, and knowing where to look often required deep knowledge of the service. Relevant signals were easy to miss. When we encountered the same kinds of problems repeatedly—like regressions in ChatGPT’s first token latency—we built advanced tools, like CPU and allocation profilers, to make those investigations easier. Over time, we accumulated more than 30 performance tools.2. Codex as a partner. Once we had Codex as a partner, engineers could ask, “Why did this get slower?” and work with Codex to investigate, even in an unfamiliar system. Codex could search large codebases, find relevant metrics and traces, follow interactions across services, and do appropriate profiling to piece together what had happened. But engineers still had to steer the process, knowing which performance tools to use and how to guide the investigation. Difficult investigations were more debuggable for more people, with less expertise and time required.3. Reusable skills. To automate further, we captured our performance team’s investigation process in a shared skill called perf-forensics. It tells Codex which tools are available, when to use them, and how to work through a slowdown so Codex can try to resolve the problem from a few angles:Find when the slowdown beganIdentify the affected requestsInvestigate whether deployments or configuration changes could explain itWe tested and refined the perf-forensics skill using evals based on past regressions. For the types of regressions we focused on (for example, ChatGPT latency and availability drops), we could now give Codex an alert and get back a diagnosis in a single run, without guiding it through each step.4. The human proxy. As we added and refined perf-related Codex skills like perf-forensics, we started to feel like a proxy between our alerts and Codex. An alert would fire, and an engineer would simply copy and paste it into Codex with the right skill, then paste the findings back into Slack.5. Automated investigations. Next, we moved investigations to the cloud so an alert could start a Codex session automatically. Codex investigated the slowdown and posted its findings in the Slack thread, but the automation ended with that reply. An engineer still had to pick up the diagnosis, develop a fix, and check whether it worked after deployment. As alerts accumulated, even figuring out which regressions needed attention meant sifting through Slack threads.6. The factory. We built the Performance Factory to run end-to-end agentic performance engineering workflows. It tracks what each active task needs next and assigns an agent when that step is ready. Each task keeps its state and findings, so work can pass between agents and engineers and resume after a code review or deployment. For a regression, the work starts with investigating an alert and ends with verifying recovery in production. We can turn common optimizations into workflows that run regularly across our services. A workflow that spots a potential optimization can open a task in another workflow to test and implement it.The Performance Factory system’s four partsThe Performance Factory is made up of four components: a task tracker with configurable workflows, a dispatch loop that assigns ready work to agents, an API that runs Codex in the cloud, and subscriptions that wake agents when something changes (such as a GitHub review comment posting or a deployment completing).Performance Factory consists of a task service, dispatch loop, cloud execution, and subscriptions.Task service with configurable workflowsThe task service is where engineers define workflows by adding configuration files and prompts for the agents to follow, written in Markdown. The configuration lists the states a task can move through, the allowed transitions, and which steps need an engineer. Separate prompts tell agents what to do at each step.Plain textA code change task contains a configuration file, along with Markdown prompts for implementing and reviewing.Each piece of work becomes a task in a workflow, and all workflows are visible in a UI. Engineers can see what’s open, follow progress, and share a task with a teammate who might have more context or be able to help with the next step.Dispatch loop that assigns workA dispatch loop periodically checks which tasks are ready for the next step. It assigns an agent to each task by starting a Codex turn with the instructions for the task’s current state. The agent records its progress on the task and moves it to the next state if it completes the step.Codex in the cloudOnce a task is assigned, the agent needs a place to do the work, with access to the same tools an engineer would use. We built an internal API to programmatically invoke a Codex turn in the cloud. The Codex turn runs on an environment inside our network, with access to our code repositories and the observability tools our engineers use, such as metrics, logs, traces, CPU profiles, and memory snapshots.The agent runs under a service identity, so we can scope its permissions and see which actions came from our performance agents. Moving agentic work to the cloud lets us run work in parallel without relying on engineers’ laptops being open or manually managing individual cloud development boxes.Subscriptions that wake the agent when something changesSome work has to wait for a person or a deployment; an agent shouldn’t have to keep running in the meantime. When a task needs to wait, it can subscribe to events like review comments or failed CI checks on a GitHub pull request to get notified when it’s time to pick the work back up. When an incoming event matches the subscription, the task becomes ready again and the dispatch loop can then assign an agent to continue the work.The first workflow: from alert to recoveryFinding, root causing, and solving regressions is the first workflow we built inside the Performance Factory. We call it the Regression Workflow. A performance regression moves through several stages: needs investigation, root caused, fix proposed, and resolved, so we built the automated workflow to reflect and move a task through those same stages.When one of its configured performance alerts fires, the Regression Workflow opens a task that agents pick up and move across these stages. The task records its status and key findings outside of any one agent’s context so that other agents or engineers can pick up the work. Engineers can follow the work in Slack or the Regression Workflow UI, add context, and review proposed changes. After a fix is deployed, an agent checks whether the performance has recovered to its original state before resolving the task.Performance Factory regression workflowA regression is detected.An agent finds the cause and proposes a fix.An engineer takes over and decides what ships.After deployment, the agent wakes up to check recovery.Still regressed? Back to investigation.A regression is detected.A regression is detected.PauseDetectedSystemFind root causeAgentPropose fixAgentReview & deployEngineerVerify recoveryAgentResolved Active stage: DetectedThe Regression Workflow’s steps include finding the root cause, proposing a fix, reviewing and deploying, and verifying recovery.Agents find the root causeFinding the root cause (or causes) is the first major step in the workflow. The agent starts by figuring out which metrics got worse, when they started to change, and which service requests were affected. The perf-forensics skill gives Codex access to the tools our performance engineers use and a playbook for working through those questions.An alert fires when a metric crosses a threshold, but the regression may have started hours earlier. The agent looks back through the data to find when the metric first got worse. It then compares groups of requests: did similar requests actually get slower, or were we serving more requests that were already slow? Those findings point to different causes.Once it knows when and where to look, the agent examines deployments and feature flag changes, then reads the relevant code. It checks when a change actually reached the affected requests and uses logs or traces to test whether the code could explain what happened.Adversarial review strengthens the diagnosisEarly on, agents sometimes stopped at the first plausible explanation they found, and some of those diagnoses turned out to be wrong. To catch these mistakes, we added an adversarial review: a separate Codex session that tries to disprove the proposed diagnosis. The reviewer rechecks the assumptions and tests counterfactuals. If a rollout caused the slowdown, did similar requests that weren’t exposed stay healthy? If the suspected change was rolled back, did the affected metric recover?In one investigation, an agent traced failing ChatGPT conversations to errors in image downloads and marked the cause as found. The Codex reviewer noticed that the internal service responsible for locating files had already recovered while image requests were still failing. The diagnosis did not explain the continuing failures, so the AI reviewer reopened the investigation and recorded what still needed to be established.We’ve had success breaking larger jobs into focused Codex turns. In a manual review of 100 root cause suggestions, we graded 70% as accurate, with 30% needing further refinement or not fully establishing causality. In the Regression Workflow, different turns investigate regressions, challenge diagnoses, and propose fixes to increase accuracy numbers through checks and balances.Fixing the regression is the hard partOnce the diagnosis passes adversarial review, the workflow asks Codex to implement a fix and open a pull request. An engineer reviews the PR and decides whether to merge it. We’ve put a significant amount of work into making this handoff as easy as possible. Each task has a report with an automatically generated summary of the regression, its impact, the root cause, and the proposed fix. We’ve tuned these summaries to be easy to scan, including metrics time series, CPU flame graphs, and traces to help engineers quickly understand the findings and inspect the evidence. Engineers can edit the PR directly or use the diagnosis to write their own fix. If they leave comments on the PR, the workflow wakes an agent to address them.An example task report for a regression shows key details, root cause, and proposed fix.After an engineer deploys the fix, the workflow wakes an agent to check whether performance has returned to its earlier baseline. If it has, the agent marks the task as resolved. If performance has only partially recovered, the agent resolves the original regression and opens a new task for the remaining slowdown, which may have a different cause. If there’s no improvement, the agent records what happened and sends the regression back for investigation.Getting fixes right has been harder than finding the cause. For example, we’ve seen cases where the agent-proposed changes are more complicated than necessary or introduce new issues. Similar to the root cause step, we added an adversarial review of each proposed fix to check whether they address the root cause and help catch bugs or unintended consequences. Some fixes also require higher level architectural decisions. For example, adding a cache might improve performance but introduce stale data. Changes like these benefit from guidance from an engineer who knows the service and can weigh the tradeoffs.We’re building an evaluation and improvement loop using these cases. By comparing the fixes agents propose with the changes engineers ultimately merge, we can better understand where proposals fall short, build evals, and improve the workflow. We’re also working on using synthetic performance tests so agents can measure proposed fixes and revise them before asking an engineer for review. We’re also exploring how agents could use our other agentic review tools to identify low-risk changes and eventually ship those fixes without a human in the loop.What we learnedBuilding and running the Performance Factory has taught us a lot about developing and operating agent workflows in production. Here are some lessons for teams building their own agentic workflows.Protect engineers’ attentionThe raw output from a Codex investigation can be a wall of text that’s hard to parse. It takes effort to figure out what the agent is claiming and track down the graphs to see whether the data backs it up. We want engineers to spend their attention on decisions rather than reconstructing the agent’s work.Agents can investigate more regressions than a team has time to read about. Our first version gave engineers a UI full of findings, but someone still had to decide which information mattered. One regression can also appear as several incidents across alerts, monitors, and channels. The workflow now links related incidents, groups them by the metrics teams care about, and ranks their estimated impact. Engineers can focus on adding context, weighing tradeoffs, and deciding what to ship.Evaluate changes to your agentsUpdating an investigation skill or upgrading the model can make the agent better at diagnosing some cases and worse at others, so we need to test their impact before updating them. We test two things separately: 1) how well the agent investigates and 2) how useful its summaries are to engineers.Evaluating investigationsFor investigations, we backtest against past production incidents. We turn those incidents into a suite of test cases we can rerun when we change the skill. A Codex agent attempts to investigate and resolve the incident, and we compare the agent’s behavior to the actual past resolution data. Each case starts with a fresh Codex investigator and the original signal, stripped of details that might give away the answer.Historical tests need care. Metrics and traces expire, while evidence collected after recovery can give away the cause. A test is only useful if we understand what evidence the investigator could see. The agent can inspect code and query production telemetry within set limits. It cannot use saved memories (notes from earlier investigations), read the incident writeup, or see results from previous runs. We also inspect its transcript afterward for information that might have spoiled the test.We score when the problem began, how large it was, and which requests were affected. We also evaluate how the agent uses evidence to rank possible causes and explain what remains uncertain. An agent can get the timing right and still blame the wrong deployment, so we score those separately.Evaluating agent-to-engineer handoffsWe evaluate task summaries using snapshots of saved tasks. These make particularly good backtests because summarization does not depend on external metrics or traces that expire. A separate grader checks whether an engineer unfamiliar with the system can understand what happened and why it matters, whether useful charts were kept, and whether the proposed next step is supported by the evidence.Turn recurring failures into self-repairLastly, our agents need monitoring just as much as the services they investigate. A broken profiler or a missing permission can keep an agent from verifying what caused a slowdown. Those failures can be hard to spot in the final report, so we run an agent every day to look through messages and logs from recent runs for tool failures and access errors.We use the Performance Factory to self-repair and fix the problems this check uncovers. When the same tool error blocks several investigations, the agent creates a task to fix it. If the cause is clear, an agent proposes a fix, and an engineer reviews the change and decides whether to merge it. After merging changes like this, we also use evals to check whether agents can get past the step that previously blocked them.What’s next: broader applications, proactive monitoringThe Regression Workflow is just one instance of the agent factory pattern that moves tasks through explicit states. Agents advance the task when more work is ready; separate reviewers challenge the conclusions; engineers step in when the next decision requires judgment. Other workflows can reuse the same machinery with different states, instructions, and events.From one workflow to many. The Performance Factory can coordinate agents across workflows, with agents in one workflow creating tasks for another.Currently, our Regression Workflow starts when an alert fires. However, we’re working on more proactive ways for the Performance Factory to find slowdowns and optimizations even when no alert fires. We want agents to use the large amount of fine-grained performance data we already collect to spot regressions our alerts miss or discover optimizations. By comparing different kinds of data (batches of CPU and memory profiles, heap dumps, and request traces across releases), an agent might notice a small increase in CPU for a given request. It can then open a task in the Regression Workflow to trace the increase to its cause, implement a fix, and open a PR for an engineer to review.In cases where we already know what to look for, we’re working to encode known signals into more Performance Factory workflows. Our service rightsizing skill, for example, helps agents spot autoscaling settings that keep more replicas running than a service needs. When we learn to recognize a pattern like that, we want to build and regularly run a workflow for it across all our services, constantly optimizing in a rapidly changing environment. With the Performance Factory, what an engineer learns from one slowdown can shape how we investigate and fix the next one, even months later in a different service.Next time you’re using ChatGPT and notice a slowdown, there’s a good chance the Performance Factory is already figuring out why. Today, the bottleneck may be waiting for an engineer to review and merge the fix. We’re already exploring whether agents can do this themselves. As product development keeps accelerating with Codex, performance systems need to become increasingly autonomous and self-improving to keep up. Codex already helps us build ChatGPT faster. Now it’s helping us keep ChatGPT fast.

Read More »

Energy Secretary Keeps Critical Generation Available in Mid-Atlantic

WASHINGTON—U.S. Secretary of Energy Chris Wright today issued an emergency order to address critical grid reliability issues facing the mid-Atlantic region of the United States. The order directs PJM Interconnection L.L.C. (PJM), in coordination with Constellation Energy Corporation (Constellation Energy), to ensure Units 3 and 4 of the Eddystone Generating Station in Pennsylvania remain available to operate and to employ economic dispatch to minimize costs for the American people. The units were originally slated to shut down on May 31, 2025.   “The energy sources that perform when you need them most are the most valuable,” said U.S. Secretary of Energy Chris Wright. “President Trump and the Energy Department are committed to keeping critical generation available when demand is highest, reducing the risk of blackouts and ensuring Americans have affordable, reliable, and secure power—regardless of whether the wind is blowing or the sun is shining.”  Secretary Wright first ordered that the two Eddystone Generating Station units remain online past their planned retirement date in a May 30, 2025 emergency order. Subsequent orders were issued later in 2025 and 2026. Keeping these units operational has strengthened energy security in the PJM region, ensuring reliable electricity for American families and businesses. This order is in effect beginning on November 21, 2026, through February 18, 2027. 

Read More »

PHP2Shell aftershock: 18 more PHP flaws put enterprise web admins on alert

While there may be heavy focus on the 9.2-rated CVE-2026-129289, the lesser flaws still matter, Dickson said. For instance, in one case, one capital letter turned off encryption; a URL written as HTTPS:// or FTPS:// with an uppercase S could send passwords, tokens, form data, and files in plaintext, even through a redirect, to anyone watching the network. Not quite PHP2Shell level, but still high priority An attacker who exploits a PHP code-execution flaw gains access to whatever the PHP worker can touch, “which is just about everything,” Dickson said. Successful exploitation executes code with the privileges of the PHP worker, which holds the database password, API keys, cloud credentials, and session data. For example, he described what he called an “ugly” attack path:

Read More »

F5 scrambles to fix more critical BIG-IP security flaws

CVE-2026-121805: When an iRule that processes client traffic using the PEM:subscriber command is configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-120292: When an HTTP profile and an iRule using the MR:message route command are configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-120285: When an HTTP profile and an iRule containing the PSC:ip_address command are configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-116953: When an iRule that processes client traffic using the GENERICMESSAGE:route command is configured on a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE); CVE-2026-123637: When BIG-IP APM is provisioned and a per-request policy that includes the Client Information agent and Endpoint Security agents is applied to a virtual server, undisclosed requests can cause the Traffic Management Microkernel (TMM) to terminate and may lead to remote code execution (RCE). The company said there are no mitigations for the flaws and advised customers to install the hotfixes on vulnerable devices, and to upgrade vulnerable versions without a listed fix to a patched version. In the same batch of out of band patches, F5 also addressed CVE-2026-126822, a high severity vulnerability integer underflow in BIG-IP APM SAML that could also crash TMM and lead to denial of service. This is the second critical issue in two months for APM; September’s zero-day bug was already being exploited when it was patched. F5 did not indicate that any of the current batch is as yet being exploited but given

Read More »

Hebbia sees improvements in financial research with GPT-6 Astra

Before a banker can recommend an acquisition to their client, they need to make sense of financial statements, customer contracts and management presentations. Hebbia’s Matrix helps bring that research together, linking answers to the documents behind them.Hebbia tested OpenAI’s GPT‑6 Astra to improve two parts of that work: finding the sources a question requires and building a presentation that follows the banker’s brief.“We’ve been strong partners with OpenAI over the course of Hebbia’s lifetime, and the leaps GPT-6 Astra has made in terms of quality outputs and accuracy have truly been surprising. GPT-6 Astra shines compared to any other model we’ve tested before on these two fronts.”—Kavita Anand, Member of Technical Staff, Product, HebbiaFrom source documents to presentationsIt could work like this: a banker is helping their client assess a strategic acquisition of an industrial-software company, with roughly 2,000 files to review. What makes its revenue durable? How does it compare with peers? In Hebbia’s hypothetical example, the banker could load the data room into Matrix and examine customer concentration, contract terms and churn in cited cells. Matrix could bring in peer filings and industry news to help develop the deal thesis, then turn that research into a diligence summary and valuation deck in the firm’s template.In that workflow, GPT‑6 Astra selects documents or Matrix cells to support an answer, then turns the analysis into native PowerPoint slides that follow the user’s structure and guidelines.“Traceability is really important to our users because they’re basing millions of dollars of decisions on what the AI actually comes back with,” Kavita says. “Citations help with traceability as well as giving users confidence that the AI found the right answer.”In Hebbia’s tests, GPT‑6 Astra combined broader source coverage with highly relevant citations. It scored 13.6 percentage points above the best alternative tested on citing the required types of sources, while 99.5% of the excerpts it cited were relevant to the question.Hebbia tested a prerelease build in August-September 2026, using the same financial research cases and grading criteria for each model. Domain experts wrote or audited the criteria against source documents.“When you’re looking at a data room with thousands of documents, that’s extremely important. Even a single percentage-point gain can yield a lot of marginal benefit,” Kavita says.A presentation also has to look like work the banker would send to a client.“I don’t want to create a deck and send that to my partner that looks like AI made it,” Kavita says. “I want it to be made in my firm’s branding, and I want it to look like a human-made slide.”GPT‑6 Astra led on both following the brief and presentation quality. In Hebbia’s tests, it met 97.9% of the applicable presentation requirements, putting it 14.4 percentage points ahead of the best alternative tested on that measure. Other models traded off polish and adherence to the brief; GPT‑6 Astra ranked first on both.“In the past, we’ve seen models that really shine on one or the other of these dimensions where they’re really good at craft, but the instruction following and adhering to guidelines is not strong, or vice versa. GPT‑6 Astra is the first model we’ve seen where it’s number one on both of those fronts,” Kavita says.In Hebbia, users can open a citation to see the underlying document and highlighted passage; slide footnotes lead back to the source material, too. A review could be as specific as checking a customer-concentration figure against the source table and correcting the reporting period before the deck goes to the investment committee.Hebbia wants bankers to spend less time reformatting decks and checking sources, and more time assessing a deal. Kavita describes what a better first draft could make possible:“What’s amazing about GPT-6 Astra is its ability to create really precise outputs and artifacts that adhere to user guidelines. You can imagine the hours this will save professionals when you get a deck that is closer to your final output. The overnighters that you have to pull get reduced to a few minutes of reviewing GPT-6 Astra’s output.”—Kavita Anand, Member of Technical Staff, Product, HebbiaA future role for agentsKavita sees a future role for agents that monitor information in the background and surface signals for finance professionals to assess.“Imagine getting an important signal a few hours before your competitors,” she says. “You can get started thinking about it and see and process more opportunities that a human is not manually hunting. It can unlock a lot of value for our users.”Hebbia wants to build on these results to help finance professionals assess deals with research they can trace back to its sources.OpenAI

Read More »

ExxonMobil takes FID on Angola Block 15 extension project

The Block 15 Lifespan Extension Project, about 145 km off Angola’s northern coast, will utilize existing infrastructure to develop additional resources throughout the Kizomba Hub area. The investment decision completes project FIDs following approval of Kizomba C redevelopment in 2025. ANPG also noted a hydrocarbon discovery in the block at the Tchihumba West-01ST1 exploration well, about 350 km northwest of Luanda. The discovery is the 21st in the block and the second this year. The well, drilled by the Valaris DS-9 in about 1,400 m of water, encountered 65.4 m of sandstone reservoir with hydrocarbon presence and good petrophysical properties. ExxonMobil is operator at Block 15 (36%) with partners Azule Energy Exploration (ANG) Ltd. (24%), Azule Energy ANG BV (18%), Equinor Angola (12%), and Sonongol E&P (10%).

Read More »

Combatting adversarial distillation

As frontier AI models become more capable, safeguards play an increasingly important role in how they can be used responsibly. Adversarial distillation creates a path for bad actors to reproduce these capabilities without appropriate safeguards. An actor may use a frontier model’s outputs without authorization to train another model with similar capabilities, then deploy it without comparable protections. For instance, an adversary might distill a model’s cyber capabilities and use the resulting model to help carry out cyberattacks that the original model would have refused to assist with or that its safeguards would have blocked.We’re sharing how OpenAI detects and disrupts adversarial distillation, how we’ve strengthened protections for our models’ protected reasoning, and what we’ve learned that can help strengthen defenses across the industry.Our approach to adversarial distillation combines adversarial testing, model training, classifiers, and API-level protections with investigations into coordinated attempts to extract model capabilities. These updated safeguards align more closely with those we use for biology and cybersecurity.We place particular emphasis on preventing attackers from extracting our models’ protected reasoning: the intermediate steps a model generates internally while working toward an answer, known as its chain-of-thought. These traces can contain information the final answer leaves out, making them especially valuable training data for another model.We’ve strengthened our defenses against protected reasoning extraction in GPT‑6.2 Astra and GPT‑6.2 Sol. In an internal adaptive evaluation using strong automated red-teaming, we tested whether attackers could recover protected reasoning the model had generated while answering an earlier prompt. The average attack success rate fell from 54.0% against GPT‑6 Astra with its original safeguards to 0.2% against GPT‑6.2 Astra with its updated safeguards.Why distillation is a safety issueDistillation is a common machine-learning technique in which one model learns from data produced by another. For example, a model can learn from another model’s answers, its reasoning, or its judgments about which answers are better and why. Beyond intellectual property concerns, unauthorized distillation poses serious safety risks. Through adversarial distillation, a second model can learn advanced frontier capabilities without inheriting the safeguards that govern how the original model can use them.The concern is especially clear in dual-use domains like cybersecurity and biology. A model that helps defenders find vulnerabilities in software could also help an attacker exploit them; a model that assists with biological research could also help someone develop a harmful pathogen. If such capabilities were transferred to a model with weaker safeguards, it may provide assistance that the original model would have refused, or that the original safeguards would have blocked.The importance of protecting exact reasoning traces for cyber safetyOur models help users with coding and defensive cybersecurity tasks. A second model trained on many of those benign outputs may learn skills that also apply to offensive cybersecurity tasks. To better understand this, we study how distillation, especially involving reasoning traces, may transfer risky cyber capabilities.In this setting, the model that produces the training data is called the teacher, and the model that learns from it is called the student. In an internal study on how cyber capabilities transfer, we used the content normally visible to users—final responses, reasoning summaries, and tool interactions—as our baseline for distillation, and tested how much additional cyber capability transfers when the student has access to the teacher’s exact reasoning traces.We started each training run from the same pretrained student model. We created a supervised fine-tuning dataset consisting primarily of trajectories from Astra-level teacher models working on coding and defensive cybersecurity tasks. Each trajectory includes all of the steps a model takes as it works through a task, including any tool interactions and its final response. We evaluated each student after fine-tuning and again after a lightweight round of reinforcement learning in cybersecurity and coding environments.After supervised fine-tuning, the student trained on exact reasoning traces generally outperformed the student trained on reasoning summaries. After reinforcement learning, the exact-trace student still scored higher at the largest inference budget tested on both benchmarks shown, while the summary-trained student scored higher at lower budgets.1ExploitBenchSEC-bench ProOur cyber distillation results suggest that protecting reasoning traces can limit cyber capability transfer, even when the underlying task is one allowed under our usage policies. Across many such interactions, exact reasoning traces can teach a student skills that generalize beyond the tasks the teacher originally performed. We will continue studying how much additional capability transfers across different student models and distillation training recipes.Defenses that reinforce one anotherWe defend against adversarial distillation at two complementary levels of our safeguard stack. At the request level, model behavior, classifier checks, and API controls limit reasoning trace extraction. At the campaign level, we look for patterns in request-level and account-level signals that can reveal coordinated extraction attempts. In turn, attacks uncovered in investigations become examples for training and testing our safeguards.Our defenses against adversarial distillation combine safeguards for individual requests with campaign detection and enforcement. Findings from investigations and automated red-teaming help strengthen those safeguards.Protecting individual requests and responsesDefensive training for model behaviorAdversarial training teaches models to recognize reasoning extraction attempts and refuse to disclose protected reasoning, while preserving responses to legitimate requests for explanations and assistance. Adversarial training examples come from successful attacks identified through internal automated red-teaming, investigations, and responsible disclosures of potential security weaknesses from outside researchers or industry partners.Classifier safeguardsWe use input classifiers to detect attempts to extract protected reasoning, and output classifiers to detect leaks of protected reasoning in assistant messages and tool-call arguments. These safeguards apply to both streaming and non-streaming responses.Tool-call coverage is important because a model can expose protected reasoning in a tool’s arguments without including it in its final answer. Output classifiers can block suspected leaks before that content reaches the caller.API-level protectionsAPI-level protections restrict how protected reasoning can be reused across interactions and models. Our September disclosure described attacks that moved encrypted reasoning from one conversation into another and asked the model to reveal the reasoning in user-readable form. We strengthened protections across user, workspace, organization, and model-family boundaries, including closing a pathway that could expose another user’s encrypted reasoning.With GPT‑6.2 Astra and GPT‑6.2 Sol, we tie a model’s protected reasoning to the conversation history that produced it. As new messages are added, the model can build on its earlier reasoning. If an earlier message is rewritten, our API drops the reasoning that followed it⁠(opens in a new window). We also prevent older model families with weaker safeguards from reading GPT‑6.2 Astra and Sol’s protected reasoning.Investigating and disrupting coordinated campaignsWhen our systems detect unauthorized adversarial distillation campaigns, which often include coordinated attempts to extract our models’ reasoning, we investigate the accounts responsible. When we identify violations, we enforce against the network of accounts, consistent with our policies and terms of service. Through investigations of each campaign, we generate insights that allow us to upgrade our defenses in order to better detect similar activity. We also share indicators of compromise with other model providers so they can investigate related abuse on their own services.This work extends beyond our own platform. Our consistent-standards policy establishes shared standards for safeguards and coordinated enforcement across providers offering our models. We work closely with partners where appropriate to uphold these standards and strengthen protections across deployments.Sharing relevant findings across industry and government where appropriate can help disrupt campaigns and enable model developers and providers to better identify and defend against attempts to distill frontier models.Measuring progressResistance to reasoning extractionTo measure how well our request-level safeguards work together, we test whether attackers can still extract protected reasoning in a setup that closely mirrors production deployments of our models and safeguards. In static evaluations, we replay a fixed set of previously successful attacks to see which still expose protected reasoning. In adaptive evaluations, automated AI red-teamers adjust their approach based on how the system responds, which can uncover weaknesses the static tests miss.We also vary what testers can see. Black-box testing gives them the same view an external user would have, while white-box testing gives them additional information about the system to help them develop stronger attacks. In our adaptive black-box evaluation, we tested whether attackers could recover protected reasoning the model had generated while answering an earlier request or the current one. On GPT‑6.2 Astra with its updated safeguards, the average attack success rate was below 1% for each attack type. Both rates were substantially lower than on GPT‑6 Astra with its original safeguards. We’re now applying selected safeguard improvements to GPT‑6 Astra as well.Attack success rate (ASR) in adaptive black-box red-teaming against GPT‑6 Astra and GPT‑6.2 Astra, aggregated across six attacker models. Lower ASR is better.Campaign detection and enforcementWe have seen multiple unauthorized adversarial distillation campaigns on our platform this year, and shared initial findings in our September report. Our weekly enforcement actions against these campaigns have ranged in scale from thousands to tens of thousands of accounts, with the rate increasing significantly this past summer.In order to measure the effectiveness of our efforts, we also track the prevalence of distillation abuse, as well as the time between detection and enforcement. Our objective is to constrain the volume and value of the data distillers are able to obtain, to increase the cost of distillation campaigns, and to shorten the time needed to respond to new attacks. We continue to improve these defenses through adversarial testing and investigations.The national security implications of adversarial distillationAdversarial distillation also poses a national security risk. A foreign actor that reproduces advanced capabilities from a U.S. model can deploy them without the protections its developer put in place. In February, we briefed⁠(opens in a new window) Congress on activity indicating that DeepSeek was attempting to distill U.S. frontier models. In September, we reported disrupting a campaign to extract our models’ protected reasoning and attributed a core part of that activity to individuals associated with Moonshot AI, the developer of Kimi. We plan to share more soon about recent campaigns we’ve disrupted.We support legislation that would give the U.S. government more tools to identify and sanction foreign actors that illicitly extract capabilities from American AI models. We’ve also told Congress we support the Chip Security Act, the AI OVERWATCH Act, and the MATCH Act. Stronger export controls can limit adversaries’ access to the hardware needed to train frontier models. Distillation defenses make it harder to acquire some of the same capabilities from models already trained. Both efforts help maintain a U.S. lead in frontier AI, giving developers more room to slow a release when they need time to test safeguards.How others can build on this workSupporting the safe proliferation of frontier capabilities is an important goal of our anti-distillation work. Our experience so far suggests several useful steps that other model developers and platforms can take:Protect private reasoning across the full interaction. Check assistant messages and tool-call arguments for leaks, and restrict how encrypted reasoning can be reused across conversations and models.Keep successful extraction attack logs found through red-teaming, production monitoring, or responsible disclosure as regression tests. Where appropriate, and consistent with applicable privacy policies, use those examples to improve model training and classifiers. When evaluating updated defenses, replay the original attacks and test whether an adaptive red-teamer can still find a way through those defenses.Combine request-level mitigations with campaign-level investigation and enforcement.Coordinate safeguard requirements and enforcement with deployment partners, and share expertise with other AI model developers, to ensure a high floor of distillation safeguards across the industry.We have previously also shared findings through the Frontier Model Forum⁠(opens in a new window) and appropriate government information-sharing channels to help other developers strengthen their defenses while keeping the details attackers could exploit private. We’ll continue working with researchers, developers, and partners to share what we learn about emerging attacks and effective defenses.As we make frontier AI broadly available, our defenses against adversarial distillation aim to reduce the risk that its capabilities are reproduced without adequate safeguards, supporting our mission to ensure that artificial general intelligence is deployed safely and benefits all of humanity.

Read More »

Stay Ahead with the Paperboy Newsletter

Your weekly dose of insights into AI, Bitcoin mining, Datacenters and Energy indusrty news. Spend 3-5 minutes and catch-up on 1 week of news.

Smarter with ONMINE

Streamline Your Growth with ONMINE