How to build a better AI benchmark

Stay Ahead, Stay ONMINE

How to build a better AI benchmark

It’s not easy being one of Silicon Valley’s favorite benchmarks. SWE-Bench (pronounced “swee bench”) launched in November 2024 to evaluate an AI model’s coding skill, using more than 2,000 real-world programming problems pulled from the public GitHub repositories of 12 different Python-based projects. In the months since then, it’s quickly become one of the most popular tests in AI. A SWE-Bench score has become a mainstay of major model releases from OpenAI, Anthropic, and Google—and outside of foundation models, the fine-tuners at AI firms are in constant competition to see who can rise above the pack. The top of the leaderboard is a pileup between three different fine tunings of Anthropic’s Claude Sonnet model and Amazon’s Q developer agent. Auto Code Rover—one of the Claude modifications—nabbed the number two spot in November, and was acquired just three months later. Despite all the fervor, this isn’t exactly a truthful assessment of which model is “better.” As the benchmark has gained prominence, “you start to see that people really want that top spot,” says John Yang, a researcher on the team that developed SWE-Bench at Princeton University. As a result, entrants have begun to game the system—which is pushing many others to wonder whether there’s a better way to actually measure AI achievement. Developers of these coding agents aren’t necessarily doing anything as straightforward cheating, but they’re crafting approaches that are too neatly tailored to the specifics of the benchmark. The initial SWE-Bench test set was limited to programs written in Python, which meant developers could gain an advantage by training their models exclusively on Python code. Soon, Yang noticed that high-scoring models would fail completely when tested on different programming languages—revealing an approach to the test that he describes as “gilded.” “It looks nice and shiny at first glance, but then you try to run it on a different language and the whole thing just kind of falls apart,” Yang says. “At that point, you’re not designing a software engineering agent. You’re designing to make a SWE-Bench agent, which is much less interesting.” The SWE-Bench issue is a symptom of a more sweeping—and complicated—problem in AI evaluation, and one that’s increasingly sparking heated debate: The benchmarks the industry uses to guide development are drifting further and further away from evaluating actual capabilities, calling their basic value into question. Making the situation worse, several benchmarks, most notably FrontierMath and Chatbot Arena, have recently come under heat for an alleged lack of transparency. Nevertheless, benchmarks still play a central role in model development, even if few experts are willing to take their results at face value. OpenAI cofounder Andrej Karpathy recently described the situation as “an evaluation crisis”: the industry has fewer trusted methods for measuring capabilities and no clear path to better ones. “Historically, benchmarks were the way we evaluated AI systems,” says Vanessa Parli, director of research at Stanford University’s Institute for Human-Centered AI. “Is that the way we want to evaluate systems going forward? And if it’s not, what is the way?” A growing group of academics and AI researchers are making the case that the answer is to go smaller, trading sweeping ambition for an approach inspired by the social sciences. Specifically, they want to focus more on testing validity, which for quantitative social scientists refers to how well a given questionnaire measures what it’s claiming to measure—and, more fundamentally, whether what it is measuring has a coherent definition. That could cause trouble for benchmarks assessing hazily defined concepts like “reasoning” or “scientific knowledge”—and for developers aiming to reach the much-hyped goal of artificial general intelligence—but it would put the industry on firmer ground as it looks to prove the worth of individual models. “Taking validity seriously means asking folks in academia, industry, or wherever to show that their system does what they say it does,” says Abigail Jacobs, a University of Michigan professor who is a central figure in the new push for validity. “I think it points to a weakness in the AI world if they want to back off from showing that they can support their claim.” The limits of traditional testing If AI companies have been slow to respond to the growing failure of benchmarks, it’s partially because the test-scoring approach has been so effective for so long. One of the biggest early successes of contemporary AI was the ImageNet challenge, a kind of antecedent to contemporary benchmarks. Released in 2010 as an open challenge to researchers, the database held more than 3 million images for AI systems to categorize into 1,000 different classes. Crucially, the test was completely agnostic to methods, and any successful algorithm quickly gained credibility regardless of how it worked. When an algorithm called AlexNet broke through in 2012, with a then unconventional form of GPU training, it became one of the foundational results of modern AI. Few would have guessed in advance that AlexNet’s convolutional neural nets would be the secret to unlocking image recognition—but after it scored well, no one dared dispute it. (One of AlexNet’s developers, Ilya Sutskever, would go on to cofound OpenAI.) A large part of what made this challenge so effective was that there was little practical difference between ImageNet’s object classification challenge and the actual process of asking a computer to recognize an image. Even if there were disputes about methods, no one doubted that the highest-scoring model would have an advantage when deployed in an actual image recognition system. But in the 12 years since, AI researchers have applied that same method-agnostic approach to increasingly general tasks. SWE-Bench is commonly used as a proxy for broader coding ability, while other exam-style benchmarks often stand in for reasoning ability. That broad scope makes it difficult to be rigorous about what a specific benchmark measures—which, in turn, makes it hard to use the findings responsibly. Where things break down Anka Reuel, a PhD student who has been focusing on the benchmark problem as part of her research at Stanford, has become convinced the evaluation problem is the result of this push toward generality. “We’ve moved from task-specific models to general-purpose models,” Reuel says. “It’s not about a single task anymore but a whole bunch of tasks, so evaluation becomes harder.” Like the University of Michigan’s Jacobs, Reuel thinks “the main issue with benchmarks is validity, even more than the practical implementation,” noting: “That’s where a lot of things break down.” For a task as complicated as coding, for instance, it’s nearly impossible to incorporate every possible scenario into your problem set. As a result, it’s hard to gauge whether a model is scoring better because it’s more skilled at coding or because it has more effectively manipulated the problem set. And with so much pressure on developers to achieve record scores, shortcuts are hard to resist. For developers, the hope is that success on lots of specific benchmarks will add up to a generally capable model. But the techniques of agentic AI mean a single AI system can encompass a complex array of different models, making it hard to evaluate whether improvement on a specific task will lead to generalization. “There’s just many more knobs you can turn,” says Sayash Kapoor, a computer scientist at Princeton and a prominent critic of sloppy practices in the AI industry. “When it comes to agents, they have sort of given up on the best practices for evaluation.” In a paper from last July, Kapoor called out specific issues in how AI models were approaching the WebArena benchmark, designed by Carnegie Mellon University researchers in 2024 as a test of an AI agent’s ability to traverse the web. The benchmark consists of more than 800 tasks to be performed on a set of cloned websites mimicking Reddit, Wikipedia, and others. Kapoor and his team identified an apparent hack in the winning model, called STeP. STeP included specific instructions about how Reddit structures URLs, allowing STeP models to jump directly to a given user’s profile page (a frequent element of WebArena tasks). This shortcut wasn’t exactly cheating, but Kapoor sees it as “a serious misrepresentation of how well the agent would work had it seen the tasks in WebArena for the first time.” Because the technique was successful, though, a similar policy has since been adopted by OpenAI’s web agent Operator. (“Our evaluation setting is designed to assess how well an agent can solve tasks given some instruction about website structures and task execution,” an OpenAI representative said when reached for comment. “This approach is consistent with how others have used and reported results with WebArena.” STeP did not respond to a request for comment.) Further highlighting the problem with AI benchmarks, late last month Kapoor and a team of researchers wrote a paper that revealed significant problems in Chatbot Arena, the popular crowdsourced evaluation system. According to the paper, the leaderboard was being manipulated; many top foundation models were conducting undisclosed private testing and releasing their scores selectively. Today, even ImageNet itself, the mother of all benchmarks, has started to fall victim to validity problems. A 2023 study from researchers at the University of Washington and Google Research found that when ImageNet-winning algorithms were pitted against six real-world data sets, the architecture improvement “resulted in little to no progress,” suggesting that the external validity of the test had reached its limit. Going smaller For those who believe the main problem is validity, the best fix is reconnecting benchmarks to specific tasks. As Reuel puts it, AI developers “have to resort to these high-level benchmarks that are almost meaningless for downstream consumers, because the benchmark developers can’t anticipate the downstream task anymore.” So what if there was a way to help the downstream consumers identify this gap? In November 2024, Reuel launched a public ranking project called BetterBench, which rates benchmarks on dozens of different criteria, such as whether the code has been publicly documented. But validity is a central theme, with particular criteria challenging designers to spell out what capability their benchmark is testing and how it relates to the tasks that make up the benchmark. “You need to have a structural breakdown of the capabilities,” Reuel says. “What are the actual skills you care about, and how do you operationalize them into something we can measure?” The results are surprising. One of the highest-scoring benchmarks is also the oldest: the Arcade Learning Environment (ALE), established in 2013 as a way to test models’ ability to learn how to play a library of Atari 2600 games. One of the lowest-scoring is the Massive Multitask Language Understanding (MMLU) benchmark, a widely used test for general language skills; by the standards of BetterBench, the connection between the questions and the underlying skill was too poorly defined. BetterBench hasn’t meant much for the reputations of specific benchmarks, at least not yet; MMLU is still widely used, and ALE is still marginal. But the project has succeeded in pushing validity into the broader conversation about how to fix benchmarks. In April, Reuel quietly joined a new research group hosted by Hugging Face, the University of Edinburgh, and EleutherAI, where she’ll develop her ideas on validity and AI model evaluation with other figures in the field. (An official announcement is expected later this month.) Irene Solaiman, Hugging Face’s head of global policy, says the group will focus on building valid benchmarks that go beyond measuring straightforward capabilities. “There’s just so much hunger for a good benchmark off the shelf that already works,” Solaiman says. “A lot of evaluations are trying to do too much.” Increasingly, the rest of the industry seems to agree. In a paper in March, researchers from Google, Microsoft, Anthropic, and others laid out a new framework for improving evaluations—with validity as the first step. “AI evaluation science must,” the researchers argue, “move beyond coarse grained claims of ‘general intelligence’ towards more task-specific and real-world relevant measures of progress.” Measuring the “squishy” things To help make this shift, some researchers are looking to the tools of social science. A February position paper argued that “evaluating GenAI systems is a social science measurement challenge,” specifically unpacking how the validity systems used in social measurements can be applied to AI benchmarking. The authors, largely employed by Microsoft’s research branch but joined by academics from Stanford and the University of Michigan, point to the standards that social scientists use to measure contested concepts like ideology, democracy, and media bias. Applied to AI benchmarks, those same procedures could offer a way to measure concepts like “reasoning” and “math proficiency” without slipping into hazy generalizations. In the social science literature, it’s particularly important that metrics begin with a rigorous definition of the concept measured by the test. For instance, if the test is to measure how democratic a society is, it first needs to establish a definition for a “democratic society” and then establish questions that are relevant to that definition. To apply this to a benchmark like SWE-Bench, designers would need to set aside the classic machine learning approach, which is to collect programming problems from GitHub and create a scheme to validate answers as true or false. Instead, they’d first need to define what the benchmark aims to measure (“ability to resolve flagged issues in software,” for instance), break that into subskills (different types of problems or types of program that the AI model can successfully process), and then finally assemble questions that accurately cover the different subskills. It’s a profound change from how AI researchers typically approach benchmarking—but for researchers like Jacobs, a coauthor on the February paper, that’s the whole point. “There’s a mismatch between what’s happening in the tech industry and these tools from social science,” she says. “We have decades and decades of thinking about how we want to measure these squishy things about humans.” Even though the idea has made a real impact in the research world, it’s been slow to influence the way AI companies are actually using benchmarks. The last two months have seen new model releases from OpenAI, Anthropic, Google, and Meta, and all of them lean heavily on multiple-choice knowledge benchmarks like MMLU—the exact approach that validity researchers are trying to move past. After all, model releases are, for the most part, still about showing increases in general intelligence, and broad benchmarks continue to be used to back up those claims. For some observers, that’s good enough. Benchmarks, Wharton professor Ethan Mollick says, are “bad measures of things, but also they’re what we’ve got.” He adds: “At the same time, the models are getting better. A lot of sins are forgiven by fast progress.” For now, the industry’s long-standing focus on artificial general intelligence seems to be crowding out a more focused validity-based approach. As long as AI models can keep growing in general intelligence, then specific applications don’t seem as compelling—even if that leaves practitioners relying on tools they no longer fully trust. “This is the tightrope we’re walking,” says Hugging Face’s Solaiman. “It’s too easy to throw the system out, but evaluations are really helpful in understanding our models, even with these limitations.” Russell Brandom is a freelance writer covering artificial intelligence. He lives in Brooklyn with his wife and two cats. This story was supported by a grant from the Tarbell Center for AI Journalism.

It’s not easy being one of Silicon Valley’s favorite benchmarks.

SWE-Bench (pronounced “swee bench”) launched in November 2024 to evaluate an AI model’s coding skill, using more than 2,000 real-world programming problems pulled from the public GitHub repositories of 12 different Python-based projects.

In the months since then, it’s quickly become one of the most popular tests in AI. A SWE-Bench score has become a mainstay of major model releases from OpenAI, Anthropic, and Google—and outside of foundation models, the fine-tuners at AI firms are in constant competition to see who can rise above the pack. The top of the leaderboard is a pileup between three different fine tunings of Anthropic’s Claude Sonnet model and Amazon’s Q developer agent. Auto Code Rover—one of the Claude modifications—nabbed the number two spot in November, and was acquired just three months later.

Despite all the fervor, this isn’t exactly a truthful assessment of which model is “better.” As the benchmark has gained prominence, “you start to see that people really want that top spot,” says John Yang, a researcher on the team that developed SWE-Bench at Princeton University. As a result, entrants have begun to game the system—which is pushing many others to wonder whether there’s a better way to actually measure AI achievement.

Developers of these coding agents aren’t necessarily doing anything as straightforward cheating, but they’re crafting approaches that are too neatly tailored to the specifics of the benchmark. The initial SWE-Bench test set was limited to programs written in Python, which meant developers could gain an advantage by training their models exclusively on Python code. Soon, Yang noticed that high-scoring models would fail completely when tested on different programming languages—revealing an approach to the test that he describes as “gilded.”

“It looks nice and shiny at first glance, but then you try to run it on a different language and the whole thing just kind of falls apart,” Yang says. “At that point, you’re not designing a software engineering agent. You’re designing to make a SWE-Bench agent, which is much less interesting.”

The SWE-Bench issue is a symptom of a more sweeping—and complicated—problem in AI evaluation, and one that’s increasingly sparking heated debate: The benchmarks the industry uses to guide development are drifting further and further away from evaluating actual capabilities, calling their basic value into question. Making the situation worse, several benchmarks, most notably FrontierMath and Chatbot Arena, have recently come under heat for an alleged lack of transparency. Nevertheless, benchmarks still play a central role in model development, even if few experts are willing to take their results at face value. OpenAI cofounder Andrej Karpathy recently described the situation as “an evaluation crisis”: the industry has fewer trusted methods for measuring capabilities and no clear path to better ones.

“Historically, benchmarks were the way we evaluated AI systems,” says Vanessa Parli, director of research at Stanford University’s Institute for Human-Centered AI. “Is that the way we want to evaluate systems going forward? And if it’s not, what is the way?”

A growing group of academics and AI researchers are making the case that the answer is to go smaller, trading sweeping ambition for an approach inspired by the social sciences. Specifically, they want to focus more on testing validity, which for quantitative social scientists refers to how well a given questionnaire measures what it’s claiming to measure—and, more fundamentally, whether what it is measuring has a coherent definition. That could cause trouble for benchmarks assessing hazily defined concepts like “reasoning” or “scientific knowledge”—and for developers aiming to reach the much–hyped goal of artificial general intelligence—but it would put the industry on firmer ground as it looks to prove the worth of individual models.

“Taking validity seriously means asking folks in academia, industry, or wherever to show that their system does what they say it does,” says Abigail Jacobs, a University of Michigan professor who is a central figure in the new push for validity. “I think it points to a weakness in the AI world if they want to back off from showing that they can support their claim.”

The limits of traditional testing

If AI companies have been slow to respond to the growing failure of benchmarks, it’s partially because the test-scoring approach has been so effective for so long.

One of the biggest early successes of contemporary AI was the ImageNet challenge, a kind of antecedent to contemporary benchmarks. Released in 2010 as an open challenge to researchers, the database held more than 3 million images for AI systems to categorize into 1,000 different classes.

Crucially, the test was completely agnostic to methods, and any successful algorithm quickly gained credibility regardless of how it worked. When an algorithm called AlexNet broke through in 2012, with a then unconventional form of GPU training, it became one of the foundational results of modern AI. Few would have guessed in advance that AlexNet’s convolutional neural nets would be the secret to unlocking image recognition—but after it scored well, no one dared dispute it. (One of AlexNet’s developers, Ilya Sutskever, would go on to cofound OpenAI.)

A large part of what made this challenge so effective was that there was little practical difference between ImageNet’s object classification challenge and the actual process of asking a computer to recognize an image. Even if there were disputes about methods, no one doubted that the highest-scoring model would have an advantage when deployed in an actual image recognition system.

But in the 12 years since, AI researchers have applied that same method-agnostic approach to increasingly general tasks. SWE-Bench is commonly used as a proxy for broader coding ability, while other exam-style benchmarks often stand in for reasoning ability. That broad scope makes it difficult to be rigorous about what a specific benchmark measures—which, in turn, makes it hard to use the findings responsibly.

Where things break down

Anka Reuel, a PhD student who has been focusing on the benchmark problem as part of her research at Stanford, has become convinced the evaluation problem is the result of this push toward generality. “We’ve moved from task-specific models to general-purpose models,” Reuel says. “It’s not about a single task anymore but a whole bunch of tasks, so evaluation becomes harder.”

Like the University of Michigan’s Jacobs, Reuel thinks “the main issue with benchmarks is validity, even more than the practical implementation,” noting: “That’s where a lot of things break down.” For a task as complicated as coding, for instance, it’s nearly impossible to incorporate every possible scenario into your problem set. As a result, it’s hard to gauge whether a model is scoring better because it’s more skilled at coding or because it has more effectively manipulated the problem set. And with so much pressure on developers to achieve record scores, shortcuts are hard to resist.

For developers, the hope is that success on lots of specific benchmarks will add up to a generally capable model. But the techniques of agentic AI mean a single AI system can encompass a complex array of different models, making it hard to evaluate whether improvement on a specific task will lead to generalization. “There’s just many more knobs you can turn,” says Sayash Kapoor, a computer scientist at Princeton and a prominent critic of sloppy practices in the AI industry. “When it comes to agents, they have sort of given up on the best practices for evaluation.”

In a paper from last July, Kapoor called out specific issues in how AI models were approaching the WebArena benchmark, designed by Carnegie Mellon University researchers in 2024 as a test of an AI agent’s ability to traverse the web. The benchmark consists of more than 800 tasks to be performed on a set of cloned websites mimicking Reddit, Wikipedia, and others. Kapoor and his team identified an apparent hack in the winning model, called STeP. STeP included specific instructions about how Reddit structures URLs, allowing STeP models to jump directly to a given user’s profile page (a frequent element of WebArena tasks).

This shortcut wasn’t exactly cheating, but Kapoor sees it as “a serious misrepresentation of how well the agent would work had it seen the tasks in WebArena for the first time.” Because the technique was successful, though, a similar policy has since been adopted by OpenAI’s web agent Operator. (“Our evaluation setting is designed to assess how well an agent can solve tasks given some instruction about website structures and task execution,” an OpenAI representative said when reached for comment. “This approach is consistent with how others have used and reported results with WebArena.” STeP did not respond to a request for comment.)

Further highlighting the problem with AI benchmarks, late last month Kapoor and a team of researchers wrote a paper that revealed significant problems in Chatbot Arena, the popular crowdsourced evaluation system. According to the paper, the leaderboard was being manipulated; many top foundation models were conducting undisclosed private testing and releasing their scores selectively.

Today, even ImageNet itself, the mother of all benchmarks, has started to fall victim to validity problems. A 2023 study from researchers at the University of Washington and Google Research found that when ImageNet-winning algorithms were pitted against six real-world data sets, the architecture improvement “resulted in little to no progress,” suggesting that the external validity of the test had reached its limit.

Going smaller

For those who believe the main problem is validity, the best fix is reconnecting benchmarks to specific tasks. As Reuel puts it, AI developers “have to resort to these high-level benchmarks that are almost meaningless for downstream consumers, because the benchmark developers can’t anticipate the downstream task anymore.” So what if there was a way to help the downstream consumers identify this gap?

In November 2024, Reuel launched a public ranking project called BetterBench, which rates benchmarks on dozens of different criteria, such as whether the code has been publicly documented. But validity is a central theme, with particular criteria challenging designers to spell out what capability their benchmark is testing and how it relates to the tasks that make up the benchmark.

“You need to have a structural breakdown of the capabilities,” Reuel says. “What are the actual skills you care about, and how do you operationalize them into something we can measure?”

The results are surprising. One of the highest-scoring benchmarks is also the oldest: the Arcade Learning Environment (ALE), established in 2013 as a way to test models’ ability to learn how to play a library of Atari 2600 games. One of the lowest-scoring is the Massive Multitask Language Understanding (MMLU) benchmark, a widely used test for general language skills; by the standards of BetterBench, the connection between the questions and the underlying skill was too poorly defined.

BetterBench hasn’t meant much for the reputations of specific benchmarks, at least not yet; MMLU is still widely used, and ALE is still marginal. But the project has succeeded in pushing validity into the broader conversation about how to fix benchmarks. In April, Reuel quietly joined a new research group hosted by Hugging Face, the University of Edinburgh, and EleutherAI, where she’ll develop her ideas on validity and AI model evaluation with other figures in the field. (An official announcement is expected later this month.)

Irene Solaiman, Hugging Face’s head of global policy, says the group will focus on building valid benchmarks that go beyond measuring straightforward capabilities. “There’s just so much hunger for a good benchmark off the shelf that already works,” Solaiman says. “A lot of evaluations are trying to do too much.”

Increasingly, the rest of the industry seems to agree. In a paper in March, researchers from Google, Microsoft, Anthropic, and others laid out a new framework for improving evaluations—with validity as the first step.

“AI evaluation science must,” the researchers argue, “move beyond coarse grained claims of ‘general intelligence’ towards more task-specific and real-world relevant measures of progress.”

Measuring the “squishy” things

To help make this shift, some researchers are looking to the tools of social science. A February position paper argued that “evaluating GenAI systems is a social science measurement challenge,” specifically unpacking how the validity systems used in social measurements can be applied to AI benchmarking.

The authors, largely employed by Microsoft’s research branch but joined by academics from Stanford and the University of Michigan, point to the standards that social scientists use to measure contested concepts like ideology, democracy, and media bias. Applied to AI benchmarks, those same procedures could offer a way to measure concepts like “reasoning” and “math proficiency” without slipping into hazy generalizations.

In the social science literature, it’s particularly important that metrics begin with a rigorous definition of the concept measured by the test. For instance, if the test is to measure how democratic a society is, it first needs to establish a definition for a “democratic society” and then establish questions that are relevant to that definition.

To apply this to a benchmark like SWE-Bench, designers would need to set aside the classic machine learning approach, which is to collect programming problems from GitHub and create a scheme to validate answers as true or false. Instead, they’d first need to define what the benchmark aims to measure (“ability to resolve flagged issues in software,” for instance), break that into subskills (different types of problems or types of program that the AI model can successfully process), and then finally assemble questions that accurately cover the different subskills.

It’s a profound change from how AI researchers typically approach benchmarking—but for researchers like Jacobs, a coauthor on the February paper, that’s the whole point. “There’s a mismatch between what’s happening in the tech industry and these tools from social science,” she says. “We have decades and decades of thinking about how we want to measure these squishy things about humans.”

Even though the idea has made a real impact in the research world, it’s been slow to influence the way AI companies are actually using benchmarks.

The last two months have seen new model releases from OpenAI, Anthropic, Google, and Meta, and all of them lean heavily on multiple-choice knowledge benchmarks like MMLU—the exact approach that validity researchers are trying to move past. After all, model releases are, for the most part, still about showing increases in general intelligence, and broad benchmarks continue to be used to back up those claims.

For some observers, that’s good enough. Benchmarks, Wharton professor Ethan Mollick says, are “bad measures of things, but also they’re what we’ve got.” He adds: “At the same time, the models are getting better. A lot of sins are forgiven by fast progress.”

For now, the industry’s long-standing focus on artificial general intelligence seems to be crowding out a more focused validity-based approach. As long as AI models can keep growing in general intelligence, then specific applications don’t seem as compelling—even if that leaves practitioners relying on tools they no longer fully trust.

“This is the tightrope we’re walking,” says Hugging Face’s Solaiman. “It’s too easy to throw the system out, but evaluations are really helpful in understanding our models, even with these limitations.”

Russell Brandom is a freelance writer covering artificial intelligence. He lives in Brooklyn with his wife and two cats.

This story was supported by a grant from the Tarbell Center for AI Journalism.

Stay Ahead

Explore More Insights

Stay ahead with more perspectives on cutting-edge power, infrastructure, energy, bitcoin and AI solutions. Explore these articles to uncover strategies and insights shaping the future of industries.

F5 to acquire CalypsoAI for advanced AI security capabilities

CalypsoAI’s platform creates what the company calls an Inference Perimeter that protects across models, vendors, and environments. The offers several products including Inference Red Team, Inference Defend, and Inference Observe, which deliver adversarial testing, threat detection and prevention, and enterprise oversight, respectively, among other capabilities. CalypsoAI says its platform proactively

HomeLM: A foundation model for ambient AI

Capabilities of a HomeLM What makes a foundation model like HomeLM powerful is its ability to learn generalizable representations of sensor streams, allowing them to be reused, recombined and adapted across diverse tasks. This fundamentally differs from traditional signal processing and machine learning pipelines in RF sensing, which are typically

Google enlists new ‘free’ data transfer service as the EU Data Act deadline approaches

Google beyond the basic requirement When asked about the rationale behind the introduction of the new service, a Google spokesperson said that the hyperscaler was going “above and beyond what the data act requires by eliminating the cost of in-parallel processing fees for customers in the EU and UK.” In

Cisco’s Splunk embeds agentic AI into security and observability products

AI-powered observability enhancements Cisco also announced it has updated Splunk Observability to use Cisco AgenticOps, which deploys AI agents to automate telemetry collection, detect issues, identify root causes, and apply fixes. The agentic AI updates help enterprise customers automate incident detection, root-cause analysis, and routine fixes. “We are making sure

U.S. Secretary of Energy Chris Wright Delivers U.S. National Statement at the General Conference of the International Atomic Energy Agency in Vienna, Austria

VIENNA, AUSTRIA— U.S. Secretary of Energy Chris Wright today delivered the U.S. National Statement at the General Conference of the International Atomic Energy Agency (IAEA) in Vienna, Austria. Secretary Wright’s full remarks from the International Atomic Energy Agency (IAEA) General Conference are below: I am honored to represent the United States of America at the 69th IAEA General Conference. I want to thank Director General Grossi and the Secretariat for your leadership. The United States welcomes the Republic of Maldives as the newest member of the IAEA. As both a lifelong energy entrepreneur and now the U.S. Secretary of Energy, I am uniquely aware of the transformative power of energy, its ability to lift billions out of poverty, drive economic growth and expand opportunity across the globe. I am also acutely aware of the challenge our world faces today in meeting rising demand for affordable, reliable and secure energy—particularly the need for baseload electric power to drive rapid progress in Artificial Intelligence. AI is rapidly emerging as the next highly energy-intensive manufacturing industry. AI manufactures intelligence out of electricity. The nations that lead in this space will also lead transformative progress in technology, healthcare, national security and innovation across the board. The energy required to power this revolution is immense—and progress will be accelerated by rapidly unlocking and deploying commercial nuclear power. The world needs more energy to meet the AI challenge and drive human progress—and the United States is boldly leading the way. With President Trump’s leadership, we are advancing American energy policies that accelerate growth, prioritize safety and enhance global security. Earlier this year, President Trump issued four Executive Orders aimed at reinvigorating America’s nuclear energy industry by modernizing regulation, streamlining reactor testing, deploying reactors for national security, and reinvigorating the nuclear industrial base. As part of these

The hidden cost of ambiguous energy software terminology

Sneha Vasudevan is a project management lead at Uplight. In the face of rapid load growth, the electricity sector is experiencing unprecedented investment in advanced technologies as organizations try to balance reliability, affordability and decarbonization. Transformation is happening on both sides of the grid, with the scale of consumer adoption of distributed energy resources approaching that of utility-scale generation capacity. Residential customers are installing heat pumps, electric vehicles and charging equipment, solar panels, and home batteries while food corporations, logistics companies and school districts electrify their vehicle fleets and implement sophisticated energy management systems. The consumer distributed energy resource hardware investment boom is resulting in increased utility spending on sophisticated software platforms to manage thousands of independently owned energy assets. Unlike the hardware world — where there is broad agreement on technical specifications of a solar panel or EV or battery — software solutions lack definitional clarity. Terms like “virtual power plant,” “fleet energy management system,” and “distributed energy resource management system” mean different things to different vendors and utilities. Successfully adapting to load growth and DER adoption hinges on the successful, scalable deployment of these software solutions. This depends on clear, mutual understanding of requirements, capabilities and outcomes among all parties. Despite the best intentions of utilities and vendors, without definitional clarity across energy software solutions, the industry remains stuck in endless scope changes and cost overruns instead of building the grid of the future. Where the industry gets lost in translation The lack of industry-wide consensus on standardized definitions for software technologies, capabilities and associated service offerings represents more than a communications issue — it’s a major barrier to meeting the increased load demand. Without shared definitions, the industry duplicates effort, misses synergies and stalls the transition to smarter energy systems. For utilities, this creates operational blind spots where

Primorsk Port Resumes Oil Loadings After Drone Attacks

At least two tankers have completed loadings at Russia’s Primorsk, showing that the Baltic Sea port has resumed operations in the aftermath of Friday’s drone attacks on the facility by Ukraine. Two crude tankers – Walrus and Samos – completed loadings at Primorsk over the weekend, according to ship tracking data compiled by Bloomberg. Walrus has left the terminal, while Samos is still anchored although is showing Aliaga in Turkey as its final destination. A third tanker Jagger is moored at the terminal. Loadings were temporarily suspended at the facility following the attacks. Three pumping stations pushing crude to Ust-Luga, another vital export terminal in the Baltic, were also hit. Ukraine has ramped up attacks on Russia’s energy facilities in the past few weeks. Kyiv has said it aims to curtail Russia’s ability to supply fuel to its front lines, while also hurting its export revenues. Primorsk is the largest Baltic oil terminal in Russia. It loaded about 970,000 barrels a day of Urals crude in August, according to Bloomberg ship tracking data. WHAT DO YOU THINK? Generated by readers, the comments included herein do not reflect the views and opinions of Rigzone. All comments are subject to editorial review. Off-topic, inappropriate or insulting comments will be removed.

North America Adds Rigs for 2 Straight Weeks

North America added seven rigs week on week, according to Baker Hughes’ latest North America rotary rig count, which was released on September 12. The U.S. added two rigs and Canada added five rigs week on week, taking the total North America rig count up to 725, comprising 539 rigs from the U.S. and 186 rigs from Canada, the count outlined. Of the total U.S. rig count of 539, 524 rigs are categorized as land rigs, 13 are categorized as offshore rigs, and two are categorized as inland water rigs. The total U.S. rig count is made up of 416 oil rigs, 118 gas rigs, and five miscellaneous rigs, according to Baker Hughes’ count, which revealed that the U.S. total comprises 471 horizontal rigs, 56 directional rigs, and 12 vertical rigs. Week on week, the U.S. offshore and inland water rig counts remained unchanged and the country’s land rig count increased by two, Baker Hughes highlighted. The U.S. oil rig count increased by two and its gas and miscellaneous rig counts remained unchanged week on week, the count showed. The U.S. directional rig count increased by two, week on week, while its horizontal rig count increased by one and its vertical rig count declined by one during the same period, the count revealed. A major state variances subcategory included in the rig count showed that, week on week, New Mexico, Ohio, and Texas each added one rig and Oklahoma dropped one rig. A major basin variances subcategory included in Baker Hughes’ rig count showed that, week on week, the Eagle Ford basin added three rigs and the Cana Woodford and Utica basins each added one rig. Canada’s total rig count of 186 is made up of 126 oil rigs, 59 gas rigs, and one miscellaneous rig, Baker Hughes pointed out.

Baker Hughes Liquefaction Tech Picked for Rio Grande LNG Train 4

Baker Hughes Co has secured a contract from Bechtel Energy Inc to deliver the main liquefaction equipment for the fourth train of NextDecade Corp’s Rio Grande liquefied natural gas (LNG) project located at the Port of Brownsville, Texas. The new contract adds to the previous framework agreement under which Baker Hughes will deliver gas turbine and refrigerant compressor technology and contractual services agreements for Trains 4 to 8, Baker Hughes said in a media release. Baker Hughes said Train 4 will replicate technology solutions provided for the first three LNG trains. The Train 4 order consists of two Frame 7 gas turbines, recognized for their established reliability and energy efficiency, along with six centrifugal compressors, Baker Hughes said. These cutting-edge solutions provide enhanced efficiency and reduced emissions, facilitating an extra LNG capacity of around 6 million tons per annum (MTPA), Baker Hughes said. “Our selection of Baker Hughes again for the Rio Grande LNG project is a testament to its reliable technology and expertise”, Bhupesh Thakkar, Bechtel’s general manager for LNG, said. “Their equipment has consistently supported the successful development of this critical infrastructure, and we look forward to their continued contribution to the project expansion”. The Rio Grande LNG facility has approximately 48 MTPA of potential liquefaction capacity under construction or in development, according to NextDecade. Train 5 is being commercialized, and Trains 6-8 are in development with permitting underway. The site can support up to 10 liquefaction trains, potentially making Rio Grande one of the largest LNG production and export facilities in the world, the developer said. To contact the author, email [email protected] WHAT DO YOU THINK? Generated by readers, the comments included herein do not reflect the views and opinions of Rigzone. All comments are subject to editorial review. Off-topic, inappropriate or insulting comments will be removed.

Baker Hughes Secures Subsea Contract for Sakarya Gas Field

Baker Hughes Company has bagged a contract from Turkish Petroleum (TPAO) and Turkish Petroleum Offshore Technology Center (TP-OTC) to supply subsea production and intelligent completion systems for the country’s strategic Sakarya Gas Field Phase 3. Baker Hughes said in a media release that it will provide deepwater horizontal tree systems with associated subsea structures and control systems to support production at depths from 6,500 to 7,200 feet. The company’s advanced, intelligent upper and lower completions systems will provide enhanced, multizonal control of subsurface operations, it said. “The development of the Sakarya gas fields has transformed Turkiye’s energy sector, leading to a more prosperous, secure future for the country”, Amerino Gatti, executive vice president of Oilfield Services and Equipment at Baker Hughes, said. “By bringing to bear our unique combination of subsea and completions technologies alongside our operational expertise and subsurface insights, Baker Hughes, TPAO, and TP-OTC are able to collaboratively unlock these crucial hydrocarbons that will power Turkiye for decades to come”. Baker Hughes said it has partnered with TPAO and TP-OTC in the Sakarya Gas Field since the beginning of its development in 2022. In Phase 3, Baker Hughes said it will combine its completions technologies, such as the InForce HCMTM-A interval control valves, SureTREAT chemical injection valves, SureSENS QPT ELITE gauges, REACH subsurface safety valves, and the SC-XP Select Zero Loss stack-pack system, with subsea production systems to enhance engineering and operational efficiencies. The energy tech company stated that deliveries and execution supporting Sakarya Gas Field Phase 3 will commence in late 2025. To contact the author, email [email protected] WHAT DO YOU THINK? Generated by readers, the comments included herein do not reflect the views and opinions of Rigzone. All comments are subject to editorial review. Off-topic, inappropriate or insulting comments will be removed. MORE FROM THIS AUTHOR

There are 121 AI processor companies. How many will succeed?

The US currently leads in AI hardware and software, but China’s DeepSeek and Huawei continue to push advanced chips, India has announced an indigenous GPU program targeting production by 2029, and policy shifts in Washington are reshaping the playing field. In Q2, the rollback of export restrictions allowed US companies like Nvidia and AMD to strike multibillion-dollar deals in Saudi Arabia. JPR categorizes vendors into five segments: IoT (ultra-low-power inference in microcontrollers or small SoCs); Edge (on-device or near-device inference in 1–100W range, used outside data centers); Automotive (distinct enough to break out from Edge); data center training; and data center inference. There is some overlap between segments as many vendors play in multiple segments. Of the five categories, inference has the most startups with 90. Peddie says the inference application list is “humongous,” with everything from wearable health monitors to smart vehicle sensor arrays, to personal items in the home, and every imaginable machine in every imaginable manufacturing and production line, plus robotic box movers and surgeons. Inference also offers the most versatility. “Smart devices” in the past, like washing machines or coffee makers, could do basically one thing and couldn’t adapt to any changes. “Inference-based systems will be able to duck and weave, adjust in real time, and find alternative solutions, quickly,” said Peddie. Peddie said despite his apparent cynicism, this is an exciting time. “There are really novel ideas being tried like analog neuron processors, and in-memory processors,” he said.

Data Center Jobs: Engineering, Construction, Commissioning, Sales, Field Service and Facility Tech Jobs Available in Major Data Center Hotspots

Each month Data Center Frontier, in partnership with Pkaza, posts some of the hottest data center career opportunities in the market. Here’s a look at some of the latest data center jobs posted on the Data Center Frontier jobs board, powered by Pkaza Critical Facilities Recruiting. Looking for Data Center Candidates? Check out Pkaza’s Active Candidate / Featured Candidate Hotlist (and coming soon free Data Center Intern listing). Data Center Critical Facility Manager Impact, TX There position is also available in: Cheyenne, WY; Ashburn, VA or Manassas, VA. This opportunity is working directly with a leading mission-critical data center developer / wholesaler / colo provider. This firm provides data center solutions custom-fit to the requirements of their client’s mission-critical operational facilities. They provide reliability of mission-critical facilities for many of the world’s largest organizations (enterprise and hyperscale customers). This career-growth minded opportunity offers exciting projects with leading-edge technology and innovation as well as competitive salaries and benefits. Electrical Commissioning Engineer New Albany, OH This traveling position is also available in: Richmond, VA; Ashburn, VA; Charlotte, NC; Atlanta, GA; Hampton, GA; Fayetteville, GA; Cedar Rapids, IA; Phoenix, AZ; Dallas, TX or Chicago, IL. *** ALSO looking for a LEAD EE and ME CxA Agents and CxA PMs. *** Our client is an engineering design and commissioning company that has a national footprint and specializes in MEP critical facilities design. They provide design, commissioning, consulting and management expertise in the critical facilities space. They have a mindset to provide reliability, energy efficiency, sustainable design and LEED expertise when providing these consulting services for enterprise, colocation and hyperscale companies. This career-growth minded opportunity offers exciting projects with leading-edge technology and innovation as well as competitive salaries and benefits. Data Center Engineering Design ManagerAshburn, VA This opportunity is working directly with a leading mission-critical data center developer /

Modernizing Legacy Data Centers for the AI Revolution with Schneider Electric’s Steven Carlini

As artificial intelligence workloads drive unprecedented compute density, the U.S. data center industry faces a formidable challenge: modernizing aging facilities that were never designed to support today’s high-density AI servers. In a recent Data Center Frontier podcast, Steven Carlini, Vice President of Innovation and Data Centers at Schneider Electric, shared his insights on how operators are confronting these transformative pressures. “Many of these data centers were built with the expectation they would go through three, four, five IT refresh cycles,” Carlini explains. “Back then, growth in rack density was moderate. Facilities were designed for 10, 12 kilowatts per rack. Now with systems like Nvidia’s Blackwell, we’re seeing 132 kilowatts per rack, and each rack can weigh 5,000 pounds.” The implications are seismic. Legacy racks, floor layouts, power distribution systems, and cooling infrastructure were simply not engineered for such extreme densities. “With densification, a lot of the power distribution, cooling systems, even the rack systems — the new servers don’t fit in those racks. You need more room behind the racks for power and cooling. Almost everything needs to be changed,” Carlini notes. For operators, the first questions are inevitably about power availability. At 132 kilowatts per rack, even a single cluster can challenge the limits of older infrastructure. Many facilities are conducting rigorous evaluations to decide whether retrofitting is feasible or whether building new sites is the more practical solution. Carlini adds, “You may have transformers spaced every hundred yards, twenty of them. Now, one larger transformer can replace that footprint, and power distribution units feed busways that supply each accelerated compute rack. The scale and complexity are unlike anything we’ve seen before.” Safety considerations also intensify with these densifications. “At 132 kilowatts, maintenance is still feasible,” Carlini says, “but as voltages rise, data centers are moving toward environments where

Google Backs Advanced Nuclear at TVA’s Clinch River as ORNL Pushes Quantum Frontiers

Inside the Hermes Reactor Design Kairos Power’s Hermes reactor is based on its KP-FHR architecture — short for fluoride salt–cooled, high-temperature reactor. Unlike conventional water-cooled reactors, Hermes uses a molten salt mixture called FLiBe (lithium fluoride and beryllium fluoride) as a coolant. Because FLiBe operates at atmospheric pressure, the design eliminates the risk of high-pressure ruptures and allows for inherently safer operation. Fuel for Hermes comes in the form of TRISO particles rather than traditional enriched uranium fuel rods. Each TRISO particle is encapsulated within ceramic layers that function like miniature containment vessels. These particles can withstand temperatures above 1,600 °C — far beyond the reactor’s normal operating range of about 700 °C. In combination with the salt coolant, Hermes achieves outlet temperatures between 650–750 °C, enabling efficient power generation and potential industrial applications such as hydrogen production. Because the salt coolant is chemically stable and requires no pressurization, the reactor can shut down and dissipate heat passively, without external power or operator intervention. This passive safety profile differentiates Hermes from traditional light-water reactors and reflects the Generation IV industry focus on safer, modular designs. From Hermes-1 to Hermes-2: Iterative Nuclear Development The first step in Kairos’ roadmap is Hermes-1, a 35 MW thermal demonstration reactor now under construction at TVA’s Clinch River site under a 2023 NRC license. Hermes-1 is not designed to generate electricity but will validate reactor physics, fuel handling, licensing strategies, and construction techniques. Building on that experience, Hermes-2 will be a 50 MW electric reactor connected to TVA’s grid, with operations targeted for 2030. Under the agreement, TVA will purchase electricity from Hermes-2 and supply it to Google’s data centers in Tennessee and Alabama. Kairos describes its development philosophy as “iterative,” scaling incrementally rather than attempting to deploy large fleets of units at once. By

NVIDIA Forecasts $3–$4 Trillion AI Market, Driving Next Wave of Infrastructure

Whenever behemoth chipmaker NVIDIA announces its quarterly earnings, those results can have a massive influence on the stock market and its position as a key indicator for the AI industry. After all, NVIDIA is the most valuable publicly traded company in the world, valued at $4.24 trillion—ahead of Microsoft ($3.74 trillion), Apple ($3.41 trillion), Alphabet, the parent company of Google ($2.57 trillion), and Amazon ($2.44 trillion). Due to its explosive growth in recent years, a single NVIDIA earnings report can move the entire market. So, when NVIDIA leaders announced during their August 27 earnings call that Q2 2026 sales surged 56% to $46.74 billion, it was a record-setting performance for the company—and investors took notice. Executive VP & CFO Colette M. Kress said the revenue exceeded leadership’s outlook as the company grew sequentially across all market platforms. She outlined a path toward substantial growth driven by AI infrastructure. Foreseeing significant long-term growth opportunities in agentic AI and considering the scale of opportunity, CEO Jensen Huang said, “Over the next 5 years, we’re going to scale into it with Blackwell [architecture for GenAI], with Rubin [successor to Blackwell], and follow-ons to scale into effectively a $3 trillion to $4 trillion AI infrastructure opportunity.” The chipmaker’s Q2 2026 earnings fell short of Wall Street’s lofty expectations, but they did demonstrate that its sales are still rising faster than those of most other tech companies. NVIDIA is expected to post revenue growth of at least 42% over the next four quarters, compared with an average of about 10% for firms in the technology-heavy Nasdaq 100 Index, according to data compiled by Bloomberg Intelligence. On August 29, two days after announcing their earnings, NVIDIA stocks slid 3% and other chip stocks also declined. This came amid a broader sell-off after server-maker Dell, a customer of those chipmakers,

Cologix and Lambda Debut NVIDIA HGX B200 AI Clusters in Columbus, Ohio

In our latest episode of the Data Center Frontier Show, we explore how powerhouse AI infrastructure is moving inland—anchored by the first NVIDIA HGX B200 cluster deployment in Columbus, Ohio. Cologix, Lambda, and Supermicro have partnered on the project, which combines Lambda’s 1-Click Clusters™, Supermicro’s energy-efficient hardware, and Cologix’s carrier-dense Scalelogix℠ COL4 facility. It’s a milestone that speaks to the rapid decentralization of AI workloads and the emergence of the Midwest as a serious player in the AI economy. Joining me for the conversation were Bill Bentley, VP Hyperscale and Cloud Sales at Cologix, and Ken Patchett, VP Data Center Infrastructure at Lambda. Why Columbus, Why Now? Asked about the significance of launching in Columbus, Patchett framed the move in terms of the coming era of “superintelligence.” “The shift to superintelligence is happening now—systems that can reason, adapt, and accelerate human progress,” Patchett said. “That requires an entirely new type of infrastructure, which means capital, vision, and the right partners. Columbus with Cologix made sense because beyond being centrally located, they’re highly connected, cost-efficient, and built to scale. We’re not chasing trends. We’re laying the groundwork for a future where intelligence infrastructure is as ubiquitous as electricity.” Bentley pointed to the city’s underlying strengths in connectivity, incentives, and utility economics. “Columbus is uniquely situated at the intersection of long-haul fiber,” Bentley said. “You’ve got state tax incentives, low-cost utilities, and a growing concentration of hyperscalers and local enterprises. The ecosystem is ripe for growth. It’s a natural geography for AI workloads that need geographic diversity without sacrificing performance.” Shifting—or Expanding—the Map for AI The guests agreed that deployments like this don’t represent a wholesale shift away from coastal hyperscale markets, but rather the expansion of AI’s footprint across multiple geographies. “I like to think of Lambda as an AI hyperscaler,”

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.

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

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

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

Stay Ahead, Stay ONMINE