Texas regulators trim, approve $2.7B CenterPoint system resiliency plan

Stay Ahead, Stay ONMINE

Texas regulators trim, approve $2.7B CenterPoint system resiliency plan

The Public Utility Commission of Texas on Thursday approved a slimmed-down version of CenterPoint Energy’s system resiliency plan, ultimately approving about $2.7 billion for strategic hardening investments over the next three years. Texas lawmakers in 2023 required electric utilities to submit resiliency plans to improve grid reliability. Oncor Electric’s $3 billion plan was the first […]

Texas lawmakers in 2023 required electric utilities to submit resiliency plans to improve grid reliability. Oncor Electric’s $3 billion plan was the first to be approved, in November, and Entergy’s $137 million plan was approved in January.

CenterPoint’s original resiliency plan, filed in January, came in with a price tag of $5.75 billion. A June settlement cut the ask to $3.2 billion, and at the PUCT open meeting last week more was removed by commissioners.

Commissioner Courtney Hjaltman said she wanted to “remove a few measures that … seem more about replacing materials that have come to the end of their useful life,” and make changes to how the cost of some vegetation management is recovered.

Hjaltman filed a memo on Wednesday discussing her proposed reductions to the plan. She raised concerns about the level of detail in CenterPoint’s vegetation management plan and how the costs to transition from a five-year to a three-year tree-trimming cycle would be recovered. She said the utility’s proposal looked more like an “augment” to its base vegetation management budget and that “the lack of verifiable details for the individual programs leaves us unable to sufficiently distinguish the resiliency investment.”

“To be clear, I am supportive of [the vegetation management projects] but consumers deserve transparency between increased base budgeting and the unique investments made through these resiliency plans,” Hjaltman said in her memo.

The day of the open meeting, CenterPoint filed a memo responding that “increasing the regulatory asset threshold as suggested would leave the Company unable to collect $30.67 million per year in incremental vegetation management expenses required to transition to a 3-year cycle” during the resiliency plan period.

At the meeting, Hjaltman said she was unconvinced.

“I still feel that CenterPoint is using base rates to cover the cost of implementing the resiliency plan, and to me that blurs the lines of what should already be done and what is the purpose of the resiliency plan, which is the ‘above and beyond,’” she said.

The PUCT voted 3-0 to approve the plan, in line with Hjaltman’s proposed changes. It includes more than two dozen resiliency projects for the 2026-2028 period, the utility said.

”These programs will benefit our customers and communities by avoiding more than 755 million outage minutes and increasing our system’s ability to take a punch from extreme weather events and get back up more quickly after storms,” CenterPoint said in an emailed statement. “We look forward to beginning work this fall.”

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.

IBM, AMD team on quantum computing

IBM and AMD are working together to blend Big Blue’s quantum computers with the chipmaker’s CPUs, GPUs and FPGAs to build intelligent, quantum-centric, high-performance computers. The plan is to combine the power and intelligence of quantum computers with the benefits of classic computing to enable new kinds of algorithms that

Spotlight report: IT careers in the AI era

AI is not only changing the world at large, it’s radically changing the professional worlds of all IT workers—from developers and SOC analysts, to helpdesk staff, I&O teams, enterprise architects, and CIOs. What new skills and knowledge will you need to thrive in the age of AI? What new roles

In the enterprise IT view of AI agents, three words matter most

Reactive AI agents Starting with the reactive category is justified, because this category makes up almost half the total number of deployed AI agent applications enterprises report. Most early enterprise AI applications were chatbots aimed at pre- and post-sale support, and some of these have been gradually converting from online

VMware Explore 2025: Broadcom tackles agentic AI security challenges

The latest challenge? Model context protocol, or MCP, servers, which allow AI agents to connect to data and tools via a natural-language interface. “With VCF 9.0, we already enabled an AI agent builder service which helps enterprises to easily build AI agents,” said Prashanth Shenoy, vice president of product marketing

Occidental Closes Sale of Midland Basin Gas Gathering Assets to Enterprise

Enterprise Products Partners LP has sealed its purchase of a natural gas gathering affiliate of Occidental Petroleum Corp. operating in the Midland Basin for $580 million in cash. “The acquired assets include certain natural gas gathering systems in the Midland Basin, as well as approximately 200 miles of natural gas gathering pipelines that support Occidental’s production activities in the Midland Basin”, Houston, Texas-based Enterprise said in a statement online. “With access to more than 1,000 drillable locations, these systems will immediately expand Enterprise’s natural gas gathering footprint in the Midland Basin and provide long-term development visibility”. The sale, signed July and completed August, is part of Occidental’s divestment program aimed at keeping debt manageable following its acquisition of CrownRock LP in a $12.4 billion transaction completed last year. “Since July 2024, Occidental has repaid $7.5 billion of debt, including proceeds from non-core Delaware Basin transactions that closed in April and July, and expects to apply an additional $580 million to debt reduction upon closing of the Midland Basin gas gathering divestiture”, the Warren Buffett-backed producer said August 6 announcing the deal with Enterprise. “These transactions bring the total divestitures since the December 2023 announcement of the CrownRock acquisition to approximately $4 billion”, Occidental said. Occidental launched a $4.5-$6 billion asset sale program when it announced its merger with CrownRock December 2023. Occidental said in its annual report for 2024 it had achieved its near-term debt repayment goal of $4.5 billion in 4Q that year. At the end of 2Q 2025 it owed $433 million in current maturities from long-term debt. Occidental accrued total current liabilities of $8.56 billion as of June, according to a regulatory filing. Meanwhile it had current assets of $8.98 billion including $2.33 billion in cash and cash equivalents. Occidental reported $396 million in adjusted net profit attributable to common shareholders for

Venezuela to Deploy Ships to Oil Exporting Hub After USA Move

Venezuela will deploy vessels to a key oil-exporting hub near Colombia in response to the US’s decision to send warships to the southern Caribbean. Venezuela’s ships, which will be larger than “patrolling” vessels, are headed north of Lake Maracaibo and the Gulf of Venezuela, Defense Minister Vladimir Padrino said on social media. The vessels will be positioned in Venezuela’s oil cradle in a port key to both Chevron Corp.’s shipments to the US and PDVSA’s exports to China. The move comes after Venezuelan President Nicolas Maduro ordered the deployment of 15,000 troops, as well as surveillance drones, to the Colombian border to counter US military activities in the region. The US last week said it would send warships to waters off Venezuela to address what Washington sees as the threat from drug cartels. Chevron’s Petroboscan oil venture, in Lake Maracaibo, pumps about 100,000 barrels a day of heavy asphalt crude that the company exports from the Gulf of Venezuela. In May, about half of the company’s total output was shipped from the area to fuelmakers on the Gulf Coast, including Valero Energy Corp., Phillips 66, PBF Energy Inc., and Chevron’s Pascagoula refinery in Mississippi. In April, two out of seven ships bound for the US departed from Lake Maracaibo, according to Bloomberg 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.

Blackline Receives First Purchase Order from ADNOC

Blackline Safety Corp. said that the Abu Dhabi National Oil Company (ADNOC) placed its first purchase order under a multi-year agreement for up to 28,000 Blackline devices, along with services. The initial order includes around 1,000 G6 wearable single-gas detectors with four years of service and 1,200 Blackline location beacons to enhance location signals inside ADNOC facilities, Blackline said in a news release. The G6 device detects hazardous gas and provides real-time location data, while the beacons enhance the location accuracy, the company said. “This first order under the new agreement with ADNOC is an important milestone in our multi-year growth strategy in the Middle East,” Andrea Lamond, vice president for growth operations for Blackline. “Through our partnership with ADNOC and Al Masaood, we are investing in the region and delivering advanced connected worker technology that helps protect workers while improving operational efficiency”. Blackline said its footprint in the Middle East has “grown significantly over the past three years” with connected safety deployments across energy companies in the region. Financial and Operational Updates In the second quarter, Blackline reported total revenue of CAD 35.9 million, a 14 percent year-over-year increase. The growth was driven by a 31 percent increase in service revenue to CAD 21.9 million, “reflecting robust demand for Blackline’s connected software services,” which increased 32 percent to $19.2 million, along with rentals, which grew 20 percent to $2.7 million, the Calgary-based company said in its most recent earnings release. Second-quarter product revenue declined 5 percent year-over-year, driven by geopolitical uncertainty that delayed deals in North America and internationally, the company said. According to Blackline, most of the company’s products are United States–Mexico–Canada Agreement (USMCA) compliant and exempt from tariffs currently in place on goods shipped to the USA from Canada. As a result, Blackline “remains well-positioned to expand

OPEC+ May Unwind 1.65MM Bpd of Cuts at Next Meeting, Analyst Warns

There is an increasing risk that OPEC+ will unwind the last 1.65 million barrels per day of cuts when they meet on September 7, Bjarne Schieldrop, Chief Commodities Analyst at Skandinaviska Enskilda Banken AB (SEB), said in an oil report sent to Rigzone by the SEB team on Monday. In the report, Schieldrop outlined that the oil market “shows pockets of strength blinking here and there” and warned that “this clearly increases the chance that OPEC+ decides to unwind the remaining 1.65 million barrels per day of voluntary cuts when they meet on 7 September to discuss production in October”. Schieldrop added in the report, however, that the group may split the unwind over two or three months. “After that the group can start again with a clean slate and discuss OPEC+ wide cuts rather than voluntary cuts by a sub-group,” Shieldrop said in the report. “That paves the way for OPEC+ wide cuts into Q1-26 where a large surplus is projected unless the group kicks in with cuts,” he added. In the report, Schieldrop highlighted “quite a few pockets of strength” in the market. “The front-end of the crude oil curves are still in backwardation,” he said in the report. “High sulfur fuel oil in ARA has weakened from parity with Brent crude in May but is still only trading at a discount of $5.6 per barrel to Brent versus a more normal discount of $10 per barrel. ARA middle distillates are trading at a premium of $25 per barrel versus Brent crude versus a more normal $15-20 per barrel,” he added. “U.S. crude stocks are at the lowest seasonal level since 2018. And lastly, the Dubai sour crude marker is trading a premium to Brent crude (light sweet crude in Europe), as highlighted by Bloomberg this morning [Monday],”

OKEA Discovers More Oil in Brage Field offshore Norway

OKEA ASA said it has found more commercial quantities of petroleum in the already producing Brage field on Norway’s side of the North Sea. The Talisker exploration well yielded preliminary estimated gross recoverable resources of 16-33 million oil-equivalent barrels (MMboe). The Statfjord formation accounted for 14-26 MMboe while the Cook formation comprised 2-7 MMboe. “In addition, hydrocarbons were encountered in two thin sandstones in the Brent group which will be further appraised by the upcoming well paths, expected completed during Q4 2025”, OKEA said. Chief executive Svein J. Liknes said, “This discovery is another example of OKEA’s strategy to utilize existing infrastructure, subsurface knowledge and drilling technology to unlock more value in the Brage area, and extend the lifetime of the field”. Earlier this year OKEA made a discovery estimated to hold recoverable volumes of 0.3-2.8 MMboe along the eastern flank of Brage. The discovery was made in the southern part of the Prince prospect in wildcat well 31/4-A-23 G. Well 31/4-A-23 F, in the northern part of the Prince prospect, turned up dry, as reported by the Directorate May 28. “The licensees will now assess the deposit as part of the further development of the Brage field”, the upstream regulator said then. “The field has been in production for a long time, and work is under way to identify new methods to improve recovery”, the Directorate said. “New wells are being drilled, often combined with investigation of nearby prospects”. Brage is part of production license 055, under which OKEA is operator with a 35.2 percent stake. Lime Petroleum AS owns 33.84 percent, DNO Norge AS 14.26 percent, Petrolia NOCO AS 12.26 percent and M Vest Energy AS 4.44 percent. Granted 1979, the license is set to expire 2030, according to information on government website Norskpetroleum.no. Brage sits 10 kilometers (6.21 miles)

Qatar’s Mansour to Acquire Nearly 20 Percent of Australia’s Invictus

Australian oil and gas explorer Invictus Energy Ltd. said that it will sell a 19.9 percent stake to a Qatari investment firm, which will help fund one of its projects in Zimbabwe. Al Mansour Holdings will purchase the stake for A$37.8 million ($24.5 million), Invictus said in a statement on Wednesday. The Qatari company has pledged to provide as much as $500 million in future financing to develop the Cabora Bassa project in the southern African nation, it said. Separately, Al Mansour – set up by Qatari’s Sheikh Mansour bin Jabor bin Jassim Al Thani – and Invictus are forming a joint venture to acquire other oil and gas assets in Africa. What do you think? We’d love to hear from you, join the conversation on the Rigzone Energy Network. The Rigzone Energy Network is a new social experience created for you and all energy professionals to Speak Up about our industry, share knowledge, connect with peers and industry insiders and engage in a professional community that will empower your career in energy.

HPE extends Juniper’s Mist AI to boost data center management

Further, Aaron stated that Marvis Actions offers automated remediations for IT-approved scenarios. Using a Human-in-the-Loop (HITL) trust model, customers can develop confidence over time, giving Marvis AI Assistant permission to automatically resolve problems such as: Correcting VLAN misconfigurations Shutting down ports to resolve network loops Upgrading noncompliant devices Handling routine policy updates and firmware compliance Resolving port-stuck issues and misconfigured access points “Each action, whether initiated by IT or executed autonomously by Marvis AI Assistant, is validated post-remediation and logged in the Marvis Actions Dashboard. This maintains full auditability and HITL oversight while building trust through consistent, accurate results,” Aaron wrote. Juniper also extended Marvis further into the vendor’s Apstra data center networking environment by letting the platform have access to Apstra’s contextual graph database, which maps the components in the data center including switches, routers, servers, links, policies and services. The idea is to let the MistAI framework understand complex queries, break them into logical components, and iteratively query data sources to synthesize actionable responses, Aaron stated. “This framework currently supports nearly 300 API queries. It will expand to enable autonomous service provisioning activities, incorporate additional data sources like elastic search, and enhance feedback mechanisms for continuous learning—critical steps toward fully self-driving data centers,” Aaron wrote. In addition to the Apstra extension, Juniper is adding Marvis Minis capabilities to data center operations. Marvis Minis set up a digital twin of a customer’s network environment to simulate and test user connections, validate network configurations, and find/detect problems without users being present and without requiring any additional hardware, according to Juniper.

Vertiv launches one-day installation package for AI data center systems

Data center infrastructure vendor Vertiv has introduced Vertiv OneCore, a fully modular data center building block design supporting AI and HPC applications intended to speed deployment of compute equipment in as little as one day. OneCore is a unified “slab-up,” factory-assembled, turnkey platform that integrates the company’s power, thermal, and IT infrastructure systems within a pre-assembled shell. Slab-up is a data center design where equipment—such as server racks and cabinets—are installed directly on a solid concrete slab floor, rather than on a raised floor system. Vertiv says the design simplifies logistics, minimizes on-site labor and complexity, and supports consistent quality, cost, and schedule outcomes. Vertiv Unify, which assists Vertiv cooling, UPS, and power management equipment to connect to building and data center management, provides integrated system visibility and centralized management.

Nvidia turns to software to speed up its data center networking hardware for AI

Typically chunks of AI tasks are distributed across GPUs, which then coordinate to provide a unified output. Adaptive routing ensures the network and GPUs over long distances are in sync when running AI workloads, Shainer said. Jitter bugs “If I retransmit the packet, I create jitter, which means one GPU out of many will be delayed and all the others have to wait for that GPU to finish,” Shainer said. The congestion control improvements remove bottlenecks by balancing transmissions across switches. Nvidia tested XGS algorithms in its server hardware and measured a 1.9x improvement in GPU-to-GPU communication compared to off-the-shelf networking technology, executives said during a briefing on the technology. Cloud providers already have long-distance high-speed networks. For example, Google’s large-scale Jupiter network uses optical switching for fast communications between its AI chips, which are called TPUs. It is important to separate the physical infrastructure from the software algorithms like XGS, Shainer said.

Fluke Networks expands testing to help ease data center networking challenges

High-density fiber connections amplify contamination risks The shift toward higher-density fiber connections has significantly complicated contamination control. Modern array connectors can house up to 24 individual fibers within a single connection point. In contrast, traditional duplex connections contained just two fibers. “The slightest little bit of dust on one of those nine micron wide fibers, which, by the way, is much smaller than a human hair, the slightest little bit of dust on any one of the 24 in that connector, and it’s not going to work,” Mullins explained. The inspection and cleaning requirements extend beyond traditional fiber testing. Each kit includes cleaning and inspection capabilities. Mullins noted that many technicians take shortcuts on fiber preparation. “Cleaning and inspecting a fiber, every time you unplug it and plug it back in, adds, like another minute worth of work. But you know what? If you don’t do it, you’re gonna pay for it down the road,” he said. Cable identification a persistent challenge In addition to the new kits, Fluke Networks is also continuing to help solve other persistent networking issues. Physical cable identification continues to plague data center operations despite advances in network management and monitoring. Fluke’s solutions address this through multiple approaches. These include tone and probe technology, remote identification systems, and active network port discovery.

Cisco ties storage networking gear to IBM z17 mainframe

“IBM Z systems are mainframes known for their ability to handle massive transaction volumes, support large-scale databases, and provide unmatched levels of security and uptime,” wrote Fausto Vaninetti, a senior solutions engineer for data center technologies at Cisco, in a blog post about the news. “The newest in the IBM Z system family, IBM z17 is the first mainframe fully engineered for the AI age, unlocking expanded capabilities for enterprise-scale AI, such as large language models, generative AI, and accelerated inferencing. However, the performance of mainframe applications depends on the underlying storage infrastructure.” SANs play a critical role in ensuring fast, reliable, and secure access to data, Vaninetti wrote: “For mainframe environments, which leverage high-speed [Fibre Connection] FICON protocol, having a robust SAN fabric that supports these requirements is non-negotiable. A solution that combines high throughput, low latency, and enterprise-class resilience is vital to ensure seamless operations and meet stringent service-level agreement requirements.” According to Vaninetti, some standout features of the MDS 9000 Series for mainframe environments include:

Scaling Up: Tract’s Master-Planned Land and Infrastructure Approach to Data Center Development

With the rapid growth of physical data center infrastructure, it’s no surprise that a niche market has emerged for companies specializing in land acquisition. Reports of massive property purchases by firms planning new facilities appear almost daily—and so do accounts of the challenges developers face before the first shovel hits the ground. As parcel sizes grow and power and water demands intensify, the complexities of acquiring and preparing these sites have only increased. Tract is a leader in this space. The Denver-based company develops master-planned data center parks, with more than 25,000 acres of potential sites under its control and plans to support over 25 GW of workload capacity. To put that into perspective, 25,000 acres is roughly 40 square miles—about two-thirds the land area of Washington, D.C., or, for European readers, two-thirds the size of Liechtenstein. Building Shovel-Ready Megasites Rather than waiting for developers to come knocking, Tract takes a proactive approach, built on the core belief that the future of data center growth lies in pre-entitled, zoned, and infrastructure-ready megasites. The company works years in advance to deliver shovel-ready campuses with reliable energy, fiber connectivity, and municipal cooperation already in place. Its model emphasizes strategic land aggregation in high-growth regions, the cultivation of long-term relationships with utilities and governments, and master planning for power, cooling, transportation, and sustainability. This integrated approach positions Tract to deliver both speed and certainty to hyperscale project developers—at scale. Tract’s leadership team brings deep industry experience. Founder and Executive Chairman Grant van Rooyen previously led acquisitions and expansions at Cologix and Teraco. President Matt Spencer brings more than 35 years of telecom and infrastructure leadership, while Chief Energy Officer Nat Sahlstrom, former head of Amazon’s global energy, water, and sustainability teams, helped make Amazon the world’s largest buyer of renewable energy. Backed by

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