Google, Microsoft, and xAI Agree to Government Review of AI Models Before Public Release

Google, Microsoft, and xAI will allow the US government to review their new AI models

Google DeepMind, Microsoft, and Elon Musk's xAI have agreed to allow the US government to review new AI models before they're released to the public. In an announcement on Tuesday, the Commerce Department's Center for AI Standards and Innovation (CAISI) says it will work with the AI companies to perform "pre-deployment evaluations and targeted research to better assess frontier AI capabilities." CAISI, which started evaluating models from OpenAI and Anthropic in 2024, says it has performed 40 reviews so far. Both companies "have renegotiated their existing partnerships with the center to better align with priorities in President Donald Trump's AI Action Plan," according to Bloomberg. The White House may take these evaluations even further in the future, as a Monday report from The New York Times suggests that Trump is considering an executive order that would "bring together tech executives and government officials" to oversee new AI models. Here's what CAISI director Chris Fall had to say in the press release: Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications. These expanded industry collaborations help us scale our work in the public interest at a critical moment.

Google, Microsoft, and xAI agree to let US government test AI models before public release -- OpenAI and Anthropic also on board after renegotiating deals with Washington

All five major frontier labs now give the Commerce Department early access to unreleased AI systems. Google, Microsoft, and Elon Musk's xAI agreed today to give the U.S. Commerce Department's Center for AI Standards and Innovation (CAISI) access to their AI models before public release, Bloomberg reports. OpenAI and Anthropic, which had existing evaluation partnerships with the center dating to 2024, renegotiated their deals to align with priorities in Trump's AI Action Plan, the agency said. The agreements mean that every major U.S. frontier AI lab now participates in voluntary pre-release government evaluations. CAISI has completed more than 40 model assessments to date, including evaluations of unreleased state-of-the-art systems, according to the Commerce Department. CAISI operates within NIST and was originally established in 2023 under Biden as the AI Safety Institute. The Trump administration renamed it last June, with Commerce Secretary Howard Lutnick calling the rebrand a move away from what he called regulation "used under the guise of national security." Despite the shift in rhetoric, the center's core function has remained largely the same: evaluating frontier models for cybersecurity, biosecurity, and chemical weapons risks. "These expanded industry collaborations help us scale our work in the public interest at a critical moment," CAISI director Chris Fall said of the new agreements. Fall took over the center after Collin Burns, a former Anthropic and OpenAI researcher, was pushed out just four days into the job. The Washington Post reported last month that White House officials were concerned about Burns's Anthropic ties, given the administration's ongoing dispute with the company. Burns had relocated across the country and given up Anthropic equity to take the position. The center still lacks permanent legal standing, and some lawmakers have introduced draft legislation to codify it, but nothing has passed. Trump's AI Action Plan, announced in July last year, directs CAISI to serve as part of an "AI evaluations ecosystem" and lead national security-related model assessments. It also instructs regulators to explore using evaluations when applying existing law to AI systems. Anthropic's renegotiated deal with CAISI sits alongside a separate and hostile set of interactions with the federal government. The Pentagon designated Anthropic a supply chain risk in March after it refused to lower guardrails on autonomous weapons, though a federal judge later called that move "Orwellian." Both Defense Secretary Pete Hegseth and Trump have outlined a six-month phaseout period for government use of Anthropic's tools, and two active lawsuits remain unresolved. The new CAISI agreements also come one day after reports that the Trump administration was considering a mandatory pre-release review process for AI models via executive order, with Anthropic's Mythos model cited as the catalyst. The voluntary agreements announced Tuesday, and any potential mandatory review framework, would run in parallel, though it remains unclear how they might interact. Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

AI labs should pass safety review to get US government contracts, group says

May 11 (Reuters) - The Trump administration should screen cutting-edge artificial intelligence models for security threats before they are publicly released and withhold lucrative government contracts from those that fail review, an advocacy group told U.S. officials on Monday. The White House is grappling with the implications of Anthropic's Mythos, which could make complex cyberattacks easier and quicker to execute, posing national ⁠security risks. Americans for Responsible Innovation urged the Trump administration to develop methods to vet upcoming frontier models from larger developers for cyberattack and weapons development capabilities. Companies should have to pass the review to be eligible for government contracts, the group said in a letter to administration officials. The U.S. Center for AI Standards and Innovation already reviews ⁠some AI models through voluntary agreements with OpenAI, Anthropic, and, more recently, Google, Microsoft and xAI. CAISI should take the lead on developing mandatory requirements, and Congress should create a permanent enforcement office ⁠within the U.S. Department of Commerce to enforce the requirements, the group said. The proposed requirements would apply to companies that spend $100 ⁠million or more a year on compute to train frontier models, or that make at least $500 million ⁠in revenue annually from AI products and services. California has a similar threshold for safety reporting requirements enacted last year. Reporting by Jody Godoy in New York Editing by Rod Nickel Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Artificial Intelligence * Data Privacy * Intellectual Property * Public Policy Jody Godoy Thomson Reuters Jody Godoy reports on tech policy and antitrust enforcement, including how regulators are responding to the rise of AI. Reach her at [email protected]

Google, xAI and Microsoft agree to US national security reviews of new AI models

The US will examine the national security implications of new AI models from Google's DeepMind, Microsoft and xAI before they are released to the public, as officials seek greater oversight of the powerful technology. The Center for AI Standards and Innovation, which sits under the commerce department, on Tuesday said it had signed a deal with the tech groups to "conduct pre-deployment evaluations and targeted research". It said the reviews would enable the government "to better assess frontier AI capabilities and advance the state of AI security". The move comes as the White House considers further measures to assess advanced models before they are widely released, said people familiar with the matter. Advisers to US President Donald Trump have mulled an executive order to impose these assessments, although discussions are at an early stage. Senior US officials have been spooked by early versions of Anthropic's new Mythos model, which the company has said has a much greater ability to identify and exploit cyber security vulnerabilities. Anthropic's chief executive Dario Amodei met White House chief of staff Susie Wiles last month, in a sign of a détente between the AI lab and the White House. The start-up had been labelled a national security threat for refusing to allow the Pentagon unrestricted use of its technology. Anthropic is suing the administration over the designation. Trump later struck a conciliatory tone, telling CNBC in an interview that Anthropic was "shaping up" and "I think we will get along with them just fine". Monday's agreement is similar to one signed with Anthropic and OpenAI during Joe Biden's administration two years ago, when Caisi was known as the US Artificial Intelligence Safety Institute. Under that deal, which mirrored the UK's policy, the government could gain access to models before their public release in order to assess and mitigate safety risks. The earlier agreements have enabled more than 40 such evaluations to date, Caisi said. Researchers at the agency are routinely provided with access to new models with safeguards removed or reduced, so they can assess the tech's capabilities and risks. The evaluations focus on AI capabilities that may pose risks to national security, with particular emphasis on cyber security, biosecurity and chemical weapons. The agency also leads assessments of AI systems developed in China, and co-ordinates findings with the Pentagon, the White House and intelligence agencies. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," agency director Chris Fall said. "These expanded industry collaborations help us scale our work in the public interest at a critical moment." Last month, tech industry representatives and AI safety campaigners called on Congress to appropriate more funding to Caisi, to help "address the complex challenges presented by AI systems".

Trump administration considers mandatory pre-release vetting of AI models -- Anthropic's Mythos cited as catalyst for policy reversal

The company that wanted AI regulation under Biden may have gotten its wish, just not on its own terms. The Trump administration is said to be discussing an executive order that would establish a government review process for new AI models before they're released to the public, The New York Times has reported, citing unnamed U.S. officials. The proposed order would create an "AI working group" of tech executives and government officials to develop oversight procedures, with White House staff briefing leaders from Anthropic, Google, and OpenAI on the plans last week. These discussions, if true, would represent a sharp departure from the administration's current stance as something of a deregulatory champion -- immediately upon taking office, the Trump administration revoked a Biden-era executive order addressing AI risks. The sudden reversal coincides with a leadership vacuum in White House AI policy. David Sacks, who led the administration's deregulation push as AI czar, left the role in March, with White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent having since taken a more active role in shaping AI policy, according to The New York Times. The new approach sounds a lot like the UK's AI Security Institute model, where government bodies evaluate frontier models against safety benchmarks before and after deployment. Officials told the New York Times that the NSA, the Office of the National Cyber Director, and the Director of National Intelligence could oversee the review. Critically, the system would grant the government early access to models without blocking their release. Perhaps unsurprisingly, the catalyst for all this appears to have been Anthropic's Mythos model, which the company's marketing described as capable of finding thousands of critical software vulnerabilities and too dangerous for public release. That naturally attracted a lot of unwanted government attention at a time when the Trump administration is already locking horns with Anthropic over the collapsed $200 million Pentagon contract. The Pentagon designated Anthropic a supply chain risk after the company refused to remove guardrails on autonomous weapons and mass surveillance, though a federal judge later called that "Orwellian." The NSA has already used Mythos to assess vulnerabilities in government Microsoft software deployments, even as other agencies remain cut off from Anthropic's tools. Some analysts have questioned whether Mythos's capabilities justify Anthropic's dramatic framing, with some studies finding that cheaper models can achieve comparable results in vulnerability discovery. A White House official told The New York Times that talk of an executive order is "speculation," and that any announcement would come from Trump himself. Dean Ball, a former senior adviser on AI in the Trump administration, told the newspaper that officials are trying to avoid overregulation while keeping pace with the technology, calling it a "tricky balance." Follow Tom's Hardware on Google News, or add us as a preferred source, to get our latest news, analysis, & reviews in your feeds.

Google, Microsoft and xAI agree to provide US government with early AI model access - Engadget

A day after reporting from The New York Times said the Trump administration was considering whether to tighten its oversight of the AI industry, Google, Microsoft and xAI have signed agreements to provide the federal government with early access to their AI systems. According to the The Wall Street Journal, the Commerce Department Center for AI Standards and Innovation (CAISI) will evaluate new models the companies develop. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," CAISI director Chris Fall told The Journal. "These expanded industry collaborations help us scale our work in the public interest at a critical moment." The deal reportedly calls for Google, Microsoft and xAI to provide their models to CAISI with reduced or even disabled safeguards in order for the organization to probe them for national security-related capabilities and risks.

Trump admin. moves further into AI oversight, will test Google, Microsoft and xAI models

CAISI, which sits under the U.S. Department of Commerce, will "conduct pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security," according to a release. The announcement builds on CAISI's previous partnerships with OpenAI and Anthropic from 2024, the center said. Those agreements have been renegotiated to reflect CAISI's directives from the Commerce Secretary Howard Lutnick and America's AI Action Plan, according to the release. Beyond CAISI's announcement on Tuesday, the White House has been weighing the creation of a new AI working group that would explore potential oversight procedures, including plans to vet models before they're released to the public, CNBC confirmed. The working group would bring together a variety of tech executives and government officials, according to a source close to the discussions who asked not to be named because the details are confidential. The group may be established through an executive order. The White House told CNBC that discussion about potential executive orders is speculation, and that any policy announcement will come directly from President Donald Trump.

US to safety test new AI models from Google, Microsoft, xAI

New artificial intelligence (AI) tools and capabilities from Google, Microsoft and xAI will now be tested by the US Department of Commerce before they are released to the public. The tech firms have agreed to voluntarily submit their models for testing through Commerce's Center for AI Standards and Innovation (CAISI). The new pacts are an expansion on agreements by AI companies like OpenAI and Anthropic that were reached during the Biden Administration, and will see AI models from all of the companies evaluated for their capabilities and security. "These expanded industry collaborations help us scale our work in the public interest at a critical moment," CAISI's director Chris Fall said. Overall, the evaluations of the AI tools will cover "testing, collaborative research and best practice development related to commercial AI systems." Google's best known AI tool, through its DeepMind subsidiary, is Gemini, a chatbot that is widely available on Google products but is now also being used in US defence and military agencies. Microsoft's best known AI tool is CoPilot, while xAI's only AI product in Grok, a chatbot that has come under widespread public scrutiny for issues were it undressed people in images. On Tuesday, CASI said it has conducted 40 previous evaluations of AI tools, including evaluation and testing of certain "state-of-the-art models that remain unreleased." The centre did not specify which models have been stopped from being released to the public. Representatives of Google, Microsoft, and SpaceX, the Elon Musk company that now controls xAI, did not respond to requests for comment. Bringing in more companies for research and safety testing of commercial AI tools marks a departure for the Trump White House, which has taken a largely hands off approach to oversight or regulation of AI and technology companies. Last year, US President Donald Trump signed a string of executive orders that formed the basis of his administration's "AI Action Plan", which he said would "remove red tape and onerous regulation" around AI development and ensure that the US will "win" through advancements and control of the technology. But with the US military expanding its use of AI, and recent claims by Anthropic that it developed a model Called Mythos that is too powerful for release to the public, the White House seems to be shifting its outlook. Senior members of Trump's staff met last month with Anthropic CEO Dario Amodei, as the BBC previously reported, even as the company is mired in a lawsuit with the US Department of Defense over Anthropic's refusal to drop safety guardrails for government use of its models.

White House Considers Vetting A.I. Models Before They Are Released

Tripp Mickle and Sheera Frenkel reported from San Francisco, and Julian Barnes and Dustin Volz from Washington. President Trump, who promoted a hands-off approach to artificial intelligence and gave Silicon Valley free rein to roll out the technology, is considering the introduction of government oversight over new A.I. models, according to U.S. officials and people briefed on the deliberations. The administration is discussing an executive order to create an A.I. working group that would bring together tech executives and government officials to examine potential oversight procedures, according to U.S. officials, who declined to be identified in order to discuss deliberations over sensitive policies. Among the potential plans is a formal government review process for new A.I. models. In meetings last week, White House officials told executives from Anthropic, Google and OpenAI about some of those plans, people briefed on the conversations said. The working group is likely to consider a number of oversight approaches, officials said. But a review process could be similar to one being developed in Britain, which has assigned several government bodies to ensure that A.I. models meet certain safety standards, people in the tech industry and the administration said. The discussions signal a stark reversal in the Trump administration's approach to A.I. Since returning to office last year, Mr. Trump has been a major booster of the technology, which he has said is vital to winning the geopolitical contest against China. Among other moves, he swiftly rolled back a Biden administration regulatory process that asked A.I. developers to perform safety evaluations and report on A.I. models with potential military applications. "We're going to make this industry absolutely the top, because right now it's a beautiful baby that's born," Mr. Trump said of A.I. at an event in July. "We have to grow that baby and let that baby thrive. We can't stop it. We can't stop it with politics. We can't stop it with foolish rules and even stupid rules." Mr. Trump left room for some rules, but he added that "they have to be more brilliant than even the technology itself." That noninterventionist policy began changing last month after the start-up Anthropic announced a new A.I. model called Mythos. Mythos is so powerful at identifying security vulnerabilities in software that it could lead to a cybersecurity "reckoning," said Anthropic, which declined to release the model to the public. The White House wants to avoid any political repercussions if a devastating A.I.-enabled cyberattack were to occur, people in the tech industry and the administration said. The administration is also evaluating whether new A.I. models could yield cyber-capabilities that could be useful to the Pentagon and U.S. intelligence agencies, they said. To get ahead of models like Mythos, some officials are pushing for a review system that would give the government first access to A.I. models, but that would not block their release, people briefed on the talks said. The shift on A.I. has sowed confusion. As conversations between the White House and tech companies continue, some executives have argued that too much government oversight will slow down U.S. innovation against China, the people briefed on the discussions said. But the companies also do not agree on how the United States should move forward with potential regulation. "The technology is moving extremely fast, and there are few formal procedures, but they also don't want to overregulate," said Dean Ball, who was a senior adviser on A.I. in the Trump administration before leaving last year for the Foundation for American Innovation. "It's a tricky balance." A White House official said that discussions of any potential executive order were "speculation" and that Mr. Trump would make any policy announcement himself. The changing policy on A.I. coincides with a leadership change at the White House. In March, David Sacks, the White House A.I. czar who had spearheaded the administration's deregulation efforts, said he was leaving the role. Susie Wiles, the White House chief of staff, and Treasury Secretary Scott Bessent have stepped in to fill Mr. Sacks's position, some of the people said. Ms. Wiles and Mr. Bessent have told people outside the administration that they plan to have a bigger hand in shaping A.I. policy. But Ms. Wiles's and Mr. Bessent's plans have been complicated by a bitter dispute between the Pentagon and Anthropic. This year, the start-up and the Pentagon became embroiled in a fight over a $200 million contract and how the military should use A.I. in warfare. When the two sides failed to agree on terms, the Pentagon cut off the government's use of Anthropic's technology in March. Anthropic has since sued the government. The conflict has made it difficult for some government agencies, which had come to rely on Anthropic's technology, according to military, intelligence and other U.S. officials. Anthropic's A.I. is still being used by the military in a system known as Maven, which helps analyze intelligence and suggests targets for airstrikes in the war in Iran. The National Security Agency has also recently used Anthropic's Mythos model to assess vulnerabilities in the U.S. government's software, people with knowledge of the work said. Last month, Ms. Wiles and Mr. Bessent held a meeting at the White House with Dario Amodei, Anthropic's chief executive, with a focus on getting the company's technology back in use by the government. Both sides later described the meeting as "productive." Officials said that if the administration moves ahead with vetting A.I. models, the working group would help determine the agencies that would help with that effort. With no federal agency responsible for all government cybersecurity work, some officials said having the N.S.A., the White House Office of the National Cyber Director and the director of national intelligence oversee the model review was the best way to proceed. The working group could also look at whether there is a role for the Center for A.I. Standards and Innovation, an agency the Biden administration established to vet A.I. models that are voluntarily shared with the government. Under Mr. Trump, the organization has been sidelined, people in the industry said, even though the White House said in an A.I. policy paper that the group should play a role in assessing "the performance and reliability of A.I. systems." Any of these moves would take the administration far from a philosophy on regulation that Vice President JD Vance outlined in a speech at an international A.I. gathering in Paris last year. At the time, he warned industry and government officials that "excessive regulation of the A.I. sector could kill a transformative industry just as it's taking off." "The A.I. future is not going to be won by hand-wringing about safety," he said. "It will be won by building." Cade Metz, Kate Conger and Tyler Pager contributed reporting.

Google, Microsoft, and xAI agree to pre-release government AI model evaluations as Mythos crisis forces oversight expansion

The Mythos crisis forced the United States government to confront a question it had been avoiding: what happens when an AI model is powerful enough to threaten national security and the government has no formal mechanism to evaluate it before the public gets access? On Tuesday, the Commerce Department announced that Google, Microsoft, and xAI have agreed to give the US government pre-release access to their AI models for evaluation. They join OpenAI and Anthropic, which have been submitting models to the same office since 2024. Five companies now account for the vast majority of frontier AI development worldwide, and all five have agreed to let a single government office test their systems before deployment. The arrangement is voluntary, has no statutory basis, and gives the government no power to block a release. It is also the closest thing the United States has to an AI oversight system, and it was built in less than two years by an office with fewer than two hundred staff. The Center for AI Standards and Innovation sits within the Commerce Department's National Institute of Standards and Technology. It was established under President Biden in 2023 as the AI Safety Institute, re-established under Trump with a new name and a reorientation toward standards and national security rather than safety research. The centre has completed more than 40 evaluations of AI models, including state-of-the-art systems that have never been released to the public. Developers frequently submit versions with safety guardrails stripped back so that evaluators can probe for national security-relevant capabilities: biological weapon synthesis pathways, cyberattack automation, and autonomous agent behaviours that could be difficult to control at scale. Chris Fall now directs the centre, following the abrupt departure of Collin Burns, a former AI researcher at Anthropic who was chosen for the role but pushed out by the White House after four days. Burns had left Anthropic, given up valuable stock, and relocated across the country to take the government position. His removal, reportedly driven by his connection to a company the administration was actively fighting, illustrates the political complexity of building an oversight system for an industry where the evaluators and the evaluated come from the same talent pool. Trump's broader AI regulatory approach has prioritised federal preemption of state regulation and a light-touch posture toward industry, but the model evaluation programme represents a harder edge: the government wants to see what these systems can do before anyone else does. The new partnerships with Google, Microsoft, and xAI expand what had been a two-company arrangement into something closer to comprehensive frontier coverage. OpenAI and Anthropic have renegotiated their existing agreements to align with Trump's AI Action Plan, which directs the centre to lead national security-related model assessments and positions it as part of a broader "evaluations ecosystem." The agreements are not contracts. They are voluntary commitments that the companies can withdraw from at any time. No statute requires pre-release evaluation. No regulation gives the centre authority to delay or block deployment. The entire system depends on the AI companies deciding, for their own strategic reasons, that giving the government early access is preferable to the alternative. The alternative, from the companies' perspective, is legislation. Several draft bills would give the centre permanent statutory authority, mandatory evaluation requirements, and the power to impose conditions on deployment. The Pentagon has already demonstrated willingness to blacklist AI companies that refuse to comply with government demands, designating Anthropic a supply-chain risk after the company refused to allow its models to be used for autonomous weapons or mass domestic surveillance. The voluntary evaluation agreements are, in part, a way for the remaining companies to demonstrate cooperation before cooperation becomes compulsion. The expansion of the evaluation programme is happening against the backdrop of the Mythos crisis. Anthropic's breakthrough model, announced in April, can autonomously discover and exploit zero-day vulnerabilities in every major operating system and web browser. It has identified thousands of high-severity bugs, including vulnerabilities that existed for decades undetected. The White House has opposed Anthropic's plan to expand access to Mythos beyond its initial consortium of launch partners. The NSA is using it despite the Pentagon's blacklist of Anthropic. The EU is demanding access to Mythos for European cyber defence, arguing that the most consequential cybersecurity tool in existence cannot remain under the exclusive control of an American company that the American government has partially blacklisted. Mythos demonstrated what the evaluation programme is designed to catch: a model whose capabilities have immediate national security implications that cannot be assessed after deployment. The centre's 40-plus evaluations since 2024 presumably identified capabilities in unreleased models that informed policy decisions, but those evaluations happened under agreements with only two companies. Google's Gemini, Microsoft's models, and xAI's Grok were not subject to pre-release government review until now. The new agreements close that gap, ensuring that the next model with Mythos-level capabilities, regardless of which lab produces it, reaches government evaluators before it reaches the public. The programme's structural weakness is obvious: it depends entirely on voluntary participation. A company that discovers its model has dangerous capabilities could, legally, decline to submit it for evaluation and release it anyway. The centre has no subpoena power, no injunctive authority, and no mechanism to compel disclosure. Its leverage is reputational and political: companies that participate signal responsibility, and companies that refuse invite regulation. But that leverage assumes the government can credibly threaten legislation, and the current administration's stated preference for light-touch regulation weakens that threat. Euro-area finance ministers have discussed Anthropic's Mythos as a financial stability concern, recognising that a cybersecurity tool capable of discovering vulnerabilities in banking infrastructure has implications beyond traditional national security. The international dimension adds pressure: if the US government cannot demonstrate that it has oversight of frontier AI models developed on its soil, other governments will impose their own requirements, fragmenting the global AI market and creating compliance costs that the companies want to avoid. The voluntary evaluation programme is, in this reading, not oversight but a prophylactic against oversight: proof that the industry is cooperating, offered in exchange for continued freedom to self-govern. The Trump administration is considering an executive order that would create a formal government review process for AI models, potentially transforming what is currently voluntary into something with regulatory teeth. A working group of tech executives and government officials would design the process, with options ranging from advisory review to mandatory pre-deployment approval. The administration's challenge is that it simultaneously wants to accelerate AI development, maintain American competitive advantage over China, avoid burdening companies with regulation, and ensure that models with national security capabilities are subject to government review. These objectives are not fully compatible, and the voluntary evaluation programme is the current attempt to reconcile them. AI capabilities are advancing into specialised domains at a pace that outstrips the government's capacity to evaluate them. The centre's 200-odd staff are assessing models that are being developed by companies with tens of thousands of researchers and hundreds of billions in capital. The asymmetry is structural: the companies will always know more about their models than the evaluators do, and the evaluation will always lag behind the frontier. What the programme provides is not comprehensive oversight. It is a window, narrow and dependent on goodwill, into what the most powerful AI systems can do before the rest of the world finds out. Five companies have agreed to keep that window open. Whether the window becomes a door, with the government able to walk through and impose conditions on what it sees, depends on whether the next Mythos-level capability arrives before or after Congress decides that voluntary cooperation is no longer enough.

Trump Reportedly Considering Executive Order Aimed at Vetting New AI Models

According to an anonymously sourced story in the New York Times, the president is considering a new oversight scheme for the AI industry. Apparently this comes from people “briefed on the conversations†held at meetings last week between Anthropic, Google, and OpenAI executives, and members of the Trump Administration. Trump is reportedly mulling an executive order that would create an “A.I. working group†made of government and tech industry representatives, and this group would discuss possible oversight plans, including what the Times calls “a formal government review process for new A.I. models.†The Times’ sources apparently claim the working group itself would be the entity determining which government agencies it would be involved withâ€"a list that could include the NSA, the White House Office of the National Cyber Director, and the office of the director of national intelligence (Tulsi Gabbard, currently). There is also, it should be noted, already an entity under the aegis of the National Institute of Standards and Technology called the Center for A.I. Standards and Innovation (CAISI), created under President Biden specifically for the vetting of AI models. But it appears that CAISI’s mission was changed shortly after Trump took office. “For far too long, censorship and regulations have been used under the guise of national security. Innovators will no longer be limited by these standards. CAISI will evaluate and enhance U.S. innovation of these rapidly developing commercial AI systems while ensuring they remain secure to our national security standards,†Secretary of Commerce Howard Lutnick said at the time. Furthermore, a policy document called “A National Policy Framework for Artificial Intelligence†released by the Trump White House released less than two months ago calls for very soft regulationsâ€"an approach that clashes significantly with what Trump now seems to be considering. It mostly prevents regulations, and contains little that’s more burdensome for Big Tech than age verification requirements. In spirit, that document was the successor of Vice President J.D. Vance’s blistering speech last year at the AI Action Summit in France, the message of which was basically, AI rules; America wins at AI; and there’s nothing any of your mid-tier economies and your European nanny states can do about it. The U.S. and U.K. refused to sign a statement at that meeting. Perhaps fittingly, the Time’s sources say the vetting plan under consideration has drawn comparisons to “one being developed in Britain,†in which multiple government entities will seek to vet AI models for safety. The plan this seems to refer to is the one that appeared to form shortly after British banks and government agencies were given a preview of Anthropic’s as-yet unreleased Claude Mythos Preview modelâ€"deemed by Anthropic to be too dangerous to release, particularly around cybersecurity. Regulators at the U.K.’s National Cyber Security Centre, its Financial Conduct Authority, its Treasury, and officials from the Bank of England were scrambling to decide a course of action as of last month.

What we know about US stress tests of Google, xAI and Microsoft AI models

WASHINGTON, May 5 (Reuters) - The Trump administration on Tuesday announced it had expanded a program to give U.S. government scientists access to unreleased artificial intelligence models to conduct risk assessments to include Google's DeepMind, xAI and Microsoft. ChatGPT maker OpenAI and Claude owner Anthropic had already been voluntarily working with the U.S. Center for AI Standards and Innovation, the team of U.S. government scientists, to test unreleased models for vulnerabilities, according to the companies. Here is what we know about the reviews: WHAT RISKS ARE THE U.S. FOCUSED ON? U.S. government scientists are focused on "demonstrable risks," such as the risk that advanced models can be used to launch cyberattacks on American infrastructure, according to the CAISI website. They want to limit opportunities for U.S. adversaries to use AI to develop chemical or biological ⁠weapons, or corrupt the data used to train American AI models. WHAT WILL COMPANIES HAND OVER? OpenAI is working with the group to test GPT-5.5-Cyber, said Chris Lehane, head of global affairs at OpenAI, in a LinkedIn post, opens new tab on Tuesday. GPT-5.5-Cyber is a variant of its latest model designed for defensive cybersecurity work. Microsoft (MSFT.O), opens new tab will work with the scientists to build shared datasets and workflows to assess advanced AI models, the company said in a statement. Microsoft did not specify which models. Anthropic gave CAISI access to both publicly available and unreleased models, allowing researchers to probe for vulnerabilities in a process known as "red-teaming," or simulating the behavior of malicious actors, the company said in September. The company also gave CAISI detailed documentation on known vulnerabilities and safety mechanisms. Google DeepMind, Alphabet's (GOOGL.O), opens new tab ⁠AI research arm, will provide access to its "proprietary models" and data, a spokesperson said. xAI did not immediately respond to a request for comment from Reuters. WHAT HAS THE U.S. FOUND SO FAR? Anthropic's work with CAISI revealed that tricks such as claiming that human review had occurred, or substituting characters, could get around safety mechanisms, the company said, adding that it had patched the vulnerabilities. OpenAI said in September that it worked ⁠with CAISI to probe vulnerabilities in its ChatGPT Agent that could have allowed sophisticated actors to bypass OpenAI's cybersecurity measures. The exploit would have allowed the attacker to "remotely control the computer systems the agent could access for that session and successfully impersonate the user for other ⁠websites they'd logged into," the company said. The companies, along with Meta, Amazon (AMZN.O), opens new tab and Inflection AI, agreed, opens new tab in 2023 to allow independent experts to check their models for biosecurity and cybersecurity risks. The U.S. government scientists, organized under a different name during former U.S. ⁠President Joe Biden's tenure, also released voluntary guidelines to protect against the risk of AI models leaking private health information or producing incorrect answers. The scientists are now working on guidelines for critical infrastructure providers, such as the communications and emergency services sectors, to test their AI systems, according to their website. Reporting by Courtney Rozen; Editing by Stephen Coates Our Standards: The Thomson Reuters Trust Principles., opens new tab * Suggested Topics: * Cybersecurity * Data Privacy * Intellectual Property * Public Policy Courtney Rozen Thomson Reuters Courtney Rozen reports on the Trump administration's transformation of federal agencies and government spending. She previously worked at Bloomberg. Jody Godoy Thomson Reuters Jody Godoy reports on tech policy and antitrust enforcement, including how regulators are responding to the rise of AI. Reach her at [email protected]

Pentagon secures early AI model access from Microsoft, Google and xAI

The U.S. government is tightening its grip on advanced artificial intelligence, moving earlier in the development cycle to assess risks before public release. In a new set of agreements, Microsoft, Google, and xAI will grant federal officials early access to their latest AI models. The decision reflects growing concern in Washington that cutting-edge systems could enable cyberattacks or military misuse if left unchecked. The Commerce Department's Center for AI Standards and Innovation will evaluate these models before deployment. Officials aim to study capabilities and identify vulnerabilities early. The move builds on a 2025 pledge by the Trump administration to partner with tech companies on national security reviews. The agreements allow government researchers to test systems under controlled conditions. Companies will provide versions of their models with reduced safeguards. This setup enables deeper probing of potential misuse scenarios. Microsoft said it will collaborate with federal scientists to test systems "in ways that probe unexpected behaviors," the company said in a statement, as reported by Reuters. The company will also help develop shared datasets and testing workflows. It previously signed a similar agreement with the United Kingdom's AI Security Institute.

The White House is considering tighter regulation of new AI models - Engadget

The New York Times reports that the White House may create a new working group to oversee AI development. A federal review of new AI models ahead of their public release is being considered as a possible power for that committee, according to the publication's sources. No clear approach has been decided, but the Times suggested it could mimic what's currently happening within the UK government, where multiple layers of oversight confirm that AI models meet safety standards. (Although the UK has recently been having its own drama about AI regulation.) There's also still a chance the entire concept fizzles and comes to nothing. If an oversight group is created, it would mark quite a reversal from the hands-off attitude presented in the White House's previously introduced AI Action Plan. The plan appeared willing to offer the AI companies most of the concessions they wanted, although it did leave a lot of potential to create plenty of new problems. Regulation for a technology industry that sure does get sued a lot seems worthwhile. Whether this administration is capable of making good decisions about that regulation is a different question.

Why some AI tools are being banned by the US government -- and what it means for you

Should AI models be vetted before release? The White House thinks it might be time The era of "move fast and break things" in AI may be coming to an abrupt halt. According to a recent New York Times report, the Trump administration is reportedly preparing a landmark Executive Order that would require Big Tech to submit their most powerful models for government vetting before they are allowed to go public. This move underscores how the rules are changing and that AI is no longer seen as a regular tech tool, but a national security asset. Here's what's behind the conversation. Why the sudden change? The catalyst for this shift appears to be the recent limited release of Anthropic's Claude Mythos. While touted as a breakthrough in cybersecurity, federal officials have raised alarms about the model's "frightening" ability to autonomously discover and exploit unpatchable software vulnerabilities in critical infrastructure. According to the report, the administration's new stance is driven by three key factors: * The 'Mythos' effect: Claims that frontier models are now skilled enough to bypass traditional cyber defenses. * Domestic compute sovereignty: A push to ensure the U.S. government has priority access to the world's most powerful processing power. * The Anthropic rift: A reported fallout between the White House and Anthropic over military usage rights, leading the administration to lean more heavily on partnerships with OpenAI and Google. Inside the discussion Last week, high-ranking White House officials reportedly met with CEOs Sundar Pichai (Google), Sam Altman (OpenAI), and Dario Amodei (Anthropic) to discuss the logistics of a government-led "working group." The goal of the discussion was said to create a standardized "red-teaming" process where federal experts audit a model's capabilities before they are ever launched. The takeaway If signed, this order could slow the breakneck pace of AI innovation in ways you'll actually notice. New "Pro" and "Ultra" model updates may take longer to arrive as they move through a rigorous vetting process, finally trading speed for added safety. Supporters say that's a win for reliability, but critics warn it could give international rivals like Deepseek an edge if they face fewer restrictions. This potential shift indicates that we may be heading toward a two-tier AI world of government-certified "safe" models for businesses and institutions and a separate, less regulated lane for hobbyists and power users. Time will tell. For now, it's a tradeoff, slower progress in exchange for tighter control. Follow Tom's Guide on Google News and add us as a preferred source to get our up-to-date news, analysis, and reviews in your feeds. Subscribe to Tom's Guide on YouTube and follow us on TikTok.

Major AI players agree to give US government early AI model access

Some of the biggest AI companies have just agreed to provide the U.S. government with early access to their new AI models. And this went down just one day after a report from the New York Times detailed how the Trump administration was looking into government oversight of new AI models. According to a new report from the Wall Street Journal, three of tech's biggest AI companies -- Google, Microsoft, and xAI -- have all reached an agreement with the Trump administration to provide access to new frontier models before they are released to the public. The three companies will provide this access to the Commerce Department's Center for AI Standards and Innovation (CAISI), which will evaluate new AI models on their capabilities and security. OpenAI and Anthropic have both previously agreed to a similar agreement with the Commerce Department in 2024. CAISI has already completed over 40 evaluations on AI models before their release to the public. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," CAISI director Chris Fall said to the WSJ. "These expanded industry collaborations help us scale our work in the public interest at a critical moment." Earlier this week, the WSJ also reported that the Trump administration is looking into a "cybersecurity-focused executive order," which would create an oversight group whose role is to create standards for AI models. These recent developments come in the wake of the Trump administration's feud with AI company Anthropic earlier this year. The US government declared Anthropic and its AI chatbot Claude was a supply chain risk to national security after the AI company requested that the Trump administration not use its technology for warfare or mass surveillance purposes. Previously, the Trump administration has taken a very pro-AI stance, citing the need for U.S. companies to maintain an edge over their Chinese rivals.

US announces deals with tech firms for national security review of AI models before release

Agreements with Microsoft, Google DeepMind and xAI focus largely on recognizing cybersecurity, biosecurity and chemical weapons risks The US government has struck deals with Google DeepMind, Microsoft and xAI to review early versions of their new AI models before they are released to the public. The Center for AI Standards and Innovation (CAISI), part of the US Department of Commerce, announced the agreements on Tuesday, saying the review process would be key to understanding the capabilities of new and powerful AI models as well as to protecting US national security. These collaborations will help the federal government "scale (its) work in the public interest at a critical moment", the agency said in a press release. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," said Chris Fall, CAISI director. CAISI is an agency meant to facilitate collaboration between the tech industry and the federal government in developing standards and assessing risks for commercial AI systems. The agreement between the agency and the AI firms is focused largely on identifying national security risks tied to cybersecurity, biosecurity and chemical weapons. OpenAI and Anthropic inked similar deals with the Biden administration two years ago and CAISI notes the agency has already completed more than 40 such evaluations, including on unreleased models. It is common for developers to share unreleased AI models with the government that have reduced or removed safety guardrails, CAISI said in its press release. This helps the government "thoroughly evaluate national security-related capabilities and risks", the agency noted. The new agreements come as fears grow that the newest and most powerful AI models - such as Anthropic's Mythos - could be dangerous to release to the public; AI safety experts, government officials and tech companies fear the expansive capabilities of these models could help hackers exploit cybersecurity vulnerabilities at an unprecedented scale. Anthropic limited its rollout of Mythos to a few companies, and initiated the collaborative Project Glasswing to bring together tech companies "to secure the world's most critical software". The New York Times and Wall Street Journal reported Monday the Trump administration was mulling over a potential executive order to create a government oversight process for these AI tools; the Administration has characterized this reporting as "speculation". Google and xAI did not immediately respond to a request for comment. Microsoft announced a similar agreement in the UK on Tuesday with the government-backed AI Security Institute, which also focuses on safe AI development. "While Microsoft regularly undertakes many types of AI testing on its own, testing for national security and large-scale public safety risks necessarily must be a collaborative endeavor with governments," Microsoft wrote in a blog post about the two deals.

U.S. ramps up frontier AI testing as White House pivots toward safety

Why it matters: As AI systems grow more powerful and potentially risky, officials want to understand their security implications before they hit the market, even in a Trump administration focused on accelerating AI innovation. * It marks a sharp change from the White House's approach of prioritizing rapid innovation without guardrails in a bid to beat China. Driving the news: The announcement comes a day after reports the Trump administration is considering increased oversight of AI models via potential executive action on cybersecurity and pre-clearance of new models. * In addition to Google, Microsoft and xAI, a spokesperson said that previously announced partnerships with Anthropic and OpenAI -- first launched in 2024 -- are "ongoing and reflect updated MOUs." * Per the release, those deals "have been renegotiated" to reflect the Center for AI Standards and Innovation's directives the Commerce secretary and President Trump's AI action plan. How it works: CAISI will "conduct pre-deployment evaluations and targeted research to better assess frontier AI capabilities and advance the state of AI security," according to Commerce. * The agreements allow for government evaluations of models before public release, as well as post-deployment assessments and related research. What they're saying: "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," said CAISI director Chris Fall in a statement. * "These expanded industry collaborations help us scale our work in the public interest at a critical moment." * Fall was recently announced as director of CAISI after former Anthropic staffer Collin Burns was reportedly pushed out after just four days on the job. The big picture: Under the Biden administration, a 2023 executive order established the AI Safety Institute, which was re-named under the Trump administration.

Microsoft, xAI and Google will share AI models with US gov't for security reviews

May 5 (Reuters) - Microsoft (MSFT.O), opens new tab, Alphabet-owned (GOOGL.O), opens new tab Google and Elon Musk's xAI will give the U.S. government early access to new artificial intelligence models before their public release to allow checks for national security risks under a new deal. The Center for AI Standards and Innovation (CAISI) at the Department of Commerce said on Tuesday that the agreement would allow it to evaluate the models before deployment and conduct research to assess their capabilities and security risks. The ⁠development of advanced AI systems including Anthropic's Mythos has in recent weeks created a stir globally, including among U.S. officials and corporate America, over their ability to supercharge hackers. The statement did not mention Anthropic, which has been in a dispute with the Pentagon over guardrails on the military's use of its AI tools. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," CAISI Director Chris ⁠Fall said in a statement. CAISI, which serves as the government's main hub for AI model testing, said it has already completed more than 40 evaluations, including on cutting-edge models not yet available to the public. Developers frequently hand over ⁠versions of their models with safety guardrails stripped back so the center can probe for national security risks, the agency said. Last week, the Pentagon said it ⁠had reached agreements with seven AI companies to deploy their advanced capabilities on the Defense Department's classified networks as it seeks to ⁠broaden the range of AI providers working across the military. Microsoft, Google and xAI did not immediately respond to requests for comment. Reporting by Aditya Soni in Bengaluru; Editing by Shinjini Ganguli and Mrigank Dhaniwala Our Standards: The Thomson Reuters Trust Principles., opens new tab

US Government Will Vet Pre-Release AI Models From Google, xAI and Microsoft - Decrypt

Anthropic's new Claude Mythos model, which excels at finding cybersecurity vulnerabilities, sparked government concern. Some of the biggest players in the AI world have agreed to give the U.S. government early access to their frontier models to test them ahead of public release, with the announcement coming one day after a report that President Donald Trump's administration is weighing an executive order on the matter. On Tuesday, the Commerce Department's Center for AI Standards and Innovation announced that Google, Microsoft, and xAI have already agreed to provide the government with pre-release access to assess the systems' capabilities. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," said Chris Fall, the center's director, in a statement. "These expanded industry collaborations help us scale our work in the public interest at a critical moment." On Monday, the New York Times reported that the Trump administration is considering an executive order to create a working group that would review advanced AI models before public release. White House officials discussed the oversight plans with executives from Anthropic, Google, and OpenAI in meetings last week, per the report. The executive order discussions were prompted partly by Anthropic's announcement last month that its breakthrough Claude Mythos model was adept at finding weak points in cybersecurity defenses, raising concerns among officials about national security implications. Rather than launch Claude Mythos to the public and potentially unleash a frenzied surge to both break and fix software like web browsers and operating systems, Anthropic has instead provided access to a limited number of startups and organizations. Mozilla said it was able to find and patch 271 vulnerabilities in its Firefox web browser using Mythos. Users on Myriad -- a prediction market platform operated by Decrypt's parent company, Dastan -- don't believe that Anthropic will release Claude Mythos broadly by June 30, penciling in just a 13% chance as of this writing. Separately, the administration has clashed with Anthropic over model access. The Trump administration and Anthropic entered a contract dispute in February after Anthropic declined a request for unrestricted access to its AI models. Defense Secretary Pete Hegseth subsequently said he would designate Anthropic as a supply chain risk to national security. A federal appeals court refused to temporarily pause the designation while the lawsuit proceeds. Last week, however, Axios reported that the White House is weighing whether to reverse course and resume its partnership with Anthropic. The potential executive order marks a sharp reversal from Trump's earlier stance on AI regulation, as he's advocated for minimal oversight of the industry. "We're going to make this industry absolutely the top, because right now it's a beautiful baby that's born," he said last July. "We have to grow that baby and let that baby thrive. We can't stop it. We can't stop it with politics. We can't stop it with foolish rules and even stupid rules." Since returning to office in 2025, Trump rolled back Biden-era regulatory requirements that asked AI developers to perform safety evaluations and report on models with potential military applications. On his first day in office, he revoked a 2023 executive order signed by former President Joe Biden that required developers of AI systems posing risks to share safety test results with the government before public release.

Tech giants agree to US government AI testing programme

The US Centre for AI Standards and Innovation will evaluate models for cybersecurity, biosecurity and chemical weapons-related risks before they are released to the public. The United States government will now be testing new artificial intelligence (AI) tools before they are released to the public. Tech firms Google, Microsoft and xAI have agreed to have the US Department of Commerce examine their models through the Center for AI Standards and Innovation (CAISI). The evaluations will cover "testing, collaborative research and best practice development related to commercial AI systems," the statement reads. CAISI will also be evaluating "demonstrable risks" associated with AI systems, such as cybersecurity, biosecurity and chemical weapons risks, its website reads. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," said CAISI Director Chris Fall in a statement. "These expanded industry collaborations help us scale our work in the public interest at a critical moment. Microsoft said in a public statement that the CAISI evaluations will help them stay ahead of risks, such as AI cyber attacks, for their AI model, Copilot. Shift in emphasis for Trump The announcement that AI companies will be evaluated is a departure for President Trump, who has long argued that excessive regulation of AI systems could hurt the country's innovation and could allow China to gain an advantage. In March, Trump released his AI National Policy Framework, which says that the United States will "remove barriers to innovation" and "accelerate" the deployment of AI across various sectors. It also said that Congress will not create "any few federal rulemaking bodies to regulate AI," but instead get existing regulatory bodies and experts in specific domains to examine the models. CAISI has already conducted 40 evaluations of other models, including on some "state-of-the-art models that remain unreleased," but did not specify which models it is referring to. "Renegotiation" of existing agreements OpenAI and Anthropic signed agreements for these evaluations in 2024 under former President Joe Biden. CAISI said existing agreements had been "renegotiated," but did not elaborate on what had changed. Chris Lehane, chief global affairs officer at OpenAI, said on LinkedIn that the company provided the government with ChatGPT5.5 ahead of its public release this week to "support national security testing and evaluations." Lehane said OpenAI is working with CAISI on testing models such as GPT-5.5-Cyber, a specific model that will strengthen cyber defence capabilities and which is only available for a limited group of first users. OpenAI said it is also involved in developing a "responsible deployment strategy" for the cybersecurity model, including a playbook to distribute these models throughout the public service.

Trump considering federal AI model oversight

White House officials are exploring official government oversight of new AI models, according to the New York Times. U.S. officials, speaking on the condition of anonymity, told the publication that the Trump administration is forming an AI working group composed of tech leaders and government representatives. The group will be tasked with outlining potential oversight procedures for new models launching to market, including formal review processes, the Times reported. The proposed plans were discussed at a White House meeting last week with representatives from Anthropic, Google, and OpenAI. Potentially influenced by regulatory processes announced by UK regulators, which relegate AI oversight to relevant government bodies, the working group would also determine which U.S. agencies would be tasked with oversight. Some officials have suggested the National Security Agency (NSA), the White House Office of the National Cyber Director, and the director of national intelligence take the lead, while others have even suggested revitalizing the Biden-era Center for A.I. Standards and Innovation, according to the Times. The administration has reversed its stance on AI regulation in recent months, despite announcing a federal AI action plan that pulled back on regulation of tech companies and threatened to reduce federal funding for states that impeded AI infrastructure efforts through regulation. Trump's One Big Beautiful Bill also included limits on state governments' AI regulation, originally proposing a 10-year moratorium on state action in favor of federal oversight. Trump appointee and FCC chairman Brendan Carr has also advocated for a light-touch approach to AI regulation.

New frontier of AI forces Trump's heavy hand

Why it matters: AI has crossed a threshold that no administration -- not even one ideologically committed to staying out of its way -- can afford to ignore. * It's a sea change in both Silicon Valley and Washington, accelerated by a new class of models that can hunt down cybersecurity flaws with extraordinary speed and precision. * Anthropic's Mythos, withheld from public use due to safety concerns, was the first model to trigger panic. But with OpenAI's GPT-5.5 now matching its capabilities and Chinese labs racing to catch up, it won't be the last. The intrigue: It was only two months ago that the Pentagon declared Anthropic a "supply chain risk" and effectively blacklisted the company. * Now the White House is developing guidance that would allow agencies to get around that designation and onboard new Anthropic models, Axios's Maria Curi and Ashley Gold scooped last week. Driving the news: That's just the first step. The White House is weighing an executive order that would give the federal government a formal role in vetting all new AI models before they hit the market, the New York Times reports. * The order would create a working group of tech executives and U.S. officials to design the oversight process, with options that include a formal government review. * White House officials briefed executives from Anthropic, Google and OpenAI on early plans last week. * Some officials are pushing for a system that would give the government first access to new AI models, but would not block their release, according to the Times. In parallel, the White House's cyber office is developing an AI security framework that would require the Pentagon to safety-test AI models before they're deployed by federal, state and local governments, Axios' Sam Sabin and Ashley Gold report. * A White House official said that any policy announcement "will come directly from the president" and that discussion about "potential executive orders is speculation." Behind the scenes: Sources at top AI companies tell Axios they're cooperating with the White House's new effort. * The Trump administration recognizes the fast-growing capabilities of the models, and the labs recognize the need to partner with the government to avoid more draconian steps. * The White House push, which includes the West Wing and the National Security Council, could result in an agreement within weeks at most, according to sources involved in the conversations. * The leading labs want to work with the government to help get the cyber defensive tools into the hands of cyber defenders more quickly, the sources say. The big picture: The Trump administration spent its first year systematically dismantling every meaningful AI safety effort the Biden administration had built. * On Day 1, Trump rescinded President Biden's AI executive order, which had asked developers to perform safety evaluations and report on models with potential military applications. * Weeks later, Vice President JD Vance told the AI Action Summit in Paris that the future would be won "by building" -- not "by hand-wringing about safety." The bottom line: The White House still sees beating China in the AI race as an existential priority, and views regulation with deep skepticism.

Google, Microsoft and xAI agree to allow government safety checks of their AI models prior to release

Google, Microsoft and xAI agree to allow government safety checks of their AI models prior to release Google LLC, Microsoft Corp. and xAI have agreed to share unreleased versions of their artificial intelligence models with the U.S. Department of Commerce to ensure the technologies do not pose a threat. Today, the Center for AI Standards and Innovation, CAISI, part of the Commerce Department, announced it will take charge of testing the models in an effort to protect U.S. national security, with the onus on cybersecurity, biosecurity and chemical weapons. While the current administration has appeared to embrace a laissez-faire approach to powerful AI systems and their development, safety concerns have mounted since the partial release of Anthropic PBC's Claude Mythos model. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," said CAISI Director Chris Fall in a statement. "These expanded industry collaborations help us scale our work in the public interest at a critical moment." OpenAI Group PBC and Anthropic were the first AI companies to allow CAISI to scrutinize their models after they made an agreement with the Biden administration about two years ago. CAISI has reportedly already completed dozens of AI model evaluations, including on unreleased state-of-the-art models. The New York Times reported earlier this week that the Trump administration is discussing an executive order that will bring tech executives and government officials together in a working group where they will create a formal review process of all new AI models. This will be a volte-face for the White House, which has taken a hands-off approach to AI development as the U.S. tries to stay ahead of China. "We can't stop it with politics," Trump said last year at an AI event. "We can't stop it with foolish rules and even stupid rules." But with more public fears around AI regarding job loss, mental health and cybersecurity, it seems the government could now be moving toward a more careful approach while remaining wary of overregulation.

White house considers vetting AI Models before they are released, NYT reports

May 4 (Reuters) - U.S. President Donald Trump is considering the introduction of government oversight over new models of artificial intelligence, the New ⁠York Times reported on Monday, citing officials briefed on the deliberations. The U.S. government is discussing an executive order to create ⁠an A.I. working group that would bring together tech executives ⁠and government officials to examine potential oversight procedures, ⁠according to the newspaper. Reuters could ⁠not immediately verify the report. Reporting by Chandni Shah in Bengaluru Our Standards: The Thomson Reuters Trust Principles., opens new tab

White House explores oversight rules for new AI models

White House officials are exploring government oversight of new AI models, according to the New York Times. U.S. officials, speaking on the condition of anonymity, stated that the Trump administration is forming an AI working group composed of tech leaders and government representatives. This group will outline potential oversight procedures for new models before their market launch. Proposals for oversight were discussed at a White House meeting last week that included representatives from Anthropic, Google, and OpenAI. The working group may adopt regulatory processes similar to those announced by UK regulators, which delegate AI oversight to relevant authorities. The group will also determine which U.S. agencies will handle this oversight. Suggested agencies for leading the oversight include the National Security Agency (NSA), the White House Office of the National Cyber Director, and the director of national intelligence. Additionally, there are discussions on the potential revitalization of the Biden-era Center for A.I. Standards and Innovation, according to the Times. The administration has recently changed its approach to AI regulation. This shift contrasts with earlier announcements about a federal AI action plan that advocated for fewer restrictions on tech companies and threatened to reduce funding for states that impose restrictive AI infrastructure regulations. Trump's One Big Beautiful Bill proposed a 10-year moratorium on state action in favor of federal oversight.

AI Labs Should Pass Safety Review to Get US Government Contracts, Group Says

May 11 (Reuters) - The ⁠Trump ⁠administration should screen cutting-edge ⁠artificial intelligence models for security threats before they are publicly released and withhold lucrative government contracts from those that fail review, an advocacy group told U.S. officials on ⁠Monday. The ⁠White House is grappling with the implications of Anthropic's Mythos, which could make complex cyberattacks easier and quicker to execute, posing national security risks. Americans for Responsible Innovation urged the Trump administration to develop methods to vet upcoming ⁠frontier models ⁠from larger developers ⁠for cyberattack and weapons development capabilities. Companies should have to pass the review to be ⁠eligible for government contracts, the group said in a letter to administration officials. The U.S. Center for AI Standards and Innovation already reviews some AI models through ⁠voluntary agreements with OpenAI, Anthropic, and, more recently, Google, Microsoft ⁠and xAI. CAISI should take the lead on developing mandatory requirements, and Congress should create a permanent enforcement office within the U.S. Department of Commerce to enforce the requirements, the group said. The proposed requirements would apply to companies that spend $100 million or more a year on compute to train ⁠frontier models, or that make at least $500 million in revenue annually from AI products and services. California has a similar threshold for safety reporting requirements enacted last year. (Reporting by Jody Godoy in New YorkEditing by Rod Nickel)

Trump administration considering safety review for new AI models

Why it matters: In a post-Mythos world, the Trump administration appears to be re-evaluating its hard line against the AI safety and security measures it once shrugged off. Driving the news: The White House's Office of the National Cyber Director (ONCD) hosted two meetings last week -- one with tech and cyber companies and another with tech trade groups -- to discuss the broader security concerns raised by advanced AI models, including Anthropic's Mythos Preview, according to two sources familiar with the matter. * The office has also been discussing an AI security framework that would require the Pentagon to lead safety testing for AI deployments for federal, state and local government levels, the two sources said. * That would be an additional layer of responsibility for the government to assess the security vulnerabilities posed by a model before it's rolled out in the public sector. * The New York Times first reported Monday that the administration is considering an executive order that could charge multiple agencies with safety testing new AI models. Zoom in: Both sources said that framework is fairly far along. * One said it was in the works before the release of Mythos sparked a new cybersecurity panic. It's unclear if there will be updates to reflect advancements from Mythos and OpenAI's GPT 5.5 model. * A White House official said that any policy announcement "will come directly from the president" and that discussion about "potential executive orders is speculation." The big picture: Ever since Anthropic announced Mythos, the White House has been scrambling to understand the hacking capabilities it possesses -- and what the administration's role should be in reining in the biggest national security concerns. * The White House has also been eying executive actions that would allow federal agencies to sidestep the current ban on government agencies using Anthropic so they can use Mythos, as Axios first reported. The intrigue: While Trump immediately revoked the Biden administration's AI executive order on Day 1, many of the ideas that are now being considered already existed in that order. * Under the Biden executive order, the Commerce Department's Center for AI Standards and Innovation would have run security and safety tests to evaluate new models before they're released to the public. Reality check: Sources say that the talks at the White House remain fluid -- with specific proposals for executive action changing often. * "There are still real tensions being worked through internally," one source, who requested anonymity to discuss confidential matters, told Axios. The bottom line: Economic and tech policy voices are worried about any policy changes that could complicate deployments, while the national security community is worried about the possible of a major AI-enabled cyberattack.

Microsoft, Google, xAI giving government early access to AI models for review

Three major artificial intelligence firms have agreed to share their models with the federal government to be tested ahead of deployment, the National Institute of Standards and Technology (NIST) announced Tuesday. As part of the agreements, Google DeepMind, Microsoft and xAI will give their models to the Center for AI Standards and Innovation (CAISI) for "pre-deployment evaluations and targeted research to better assess the frontier AI capabilities and advance the state of AI security." CAISI will also conduct assessments and other research after the models are fully deployed, CAISI said, adding the AI safety office has completed more than 40 evaluations of this nature. The agreements build upon OpenAI and Anthropic's agreements in 2024, which were the first of this kind. The Hill reached out to NIST to confirm these agreements would be continuing. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," CAISI Director Chris Fall said in a statement. "These expanded industry collaborations help us scale our work in the public interest at a critical moment." The announcement comes one day after The New York Times reported the White House is considering vetting AI models before they are released. While government testing of AI models has existed in recent years, the White House's involvement would increase the oversight. The move would be a diversion for President Trump, who has long advocated for a pro-innovation, light-touch regulation approach to AI development. The Times reported the Trump administration is floating an executive order to establish an AI working group that would convene tech executives and government officials to examine "potential oversight procedures." When reached for comment on the report, a White House official said "Any policy announcement will come directly from the president. Discussion about potential executive orders is speculation." This comes against the backdrop of the Pentagon's ongoing fight with Anthropic after negotiations fell apart over safety guardrails earlier this year. The Pentagon took the unprecedented move of labeling the AI company as a supply chain risk, while Trump directed the civilian agencies to cease use of the company's products. The White House has softened this tone and hosted Anthropic leaders to learn about its latest model Mythos, which the company says is their most advanced model to date. Multiple intelligence agencies have begun testing and using the Mythos model to find security vulnerabilities. While several civilian agencies immediately began winding down Anthropic's products, a federal judge put a pause on Trump's directive, and many agencies have since revived the product.

Microsoft, xAI and Google will share AI models with US government for security reviews

Major tech firms Microsoft, Google, and xAI are partnering with the US government. They will grant early access to new artificial intelligence models. This allows for checks on national security risks before public release. The Center for AI Standards and Innovation will conduct evaluations. This initiative aims to understand advanced AI's capabilities and potential dangers. Microsoft, Alphabet-owned Google and Elon Musk's xAI will give the U.S. government early access to new artificial intelligence models before their public release to allow checks for national security risks under a new deal. Assembly Elections 2026 Election Results 2026 Live Updates: Who's ahead in which stateWest Bengal Election Results 2026 Live UpdatesTN Election Result 2026 Live Updates The Center for AI Standards and Innovation (CAISI) at the Department of Commerce said on Tuesday that the agreement would allow it to evaluate ⁠the ⁠models before deployment and conduct research to assess their capabilities and security risks. The development of advanced AI systems including Anthropic's Mythos has in recent weeks created a stir globally, including among U.S. officials and corporate America, over their ability to supercharge hackers. The ChatGPT-maker unveiled GPT-5.4-Cyber last month, a variant of its latest flagship model fine-tuned specifically for defensive cybersecurity work, following ⁠rival Anthropic's announcement of advanced AI model Mythos. Anthropic has been embroiled in a dispute with the Pentagon over guardrails on the military's ⁠use of its AI tools. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," CAISI Director Chris Fall said in a statement. The move builds on 2024 agreements with OpenAI and Anthropic, established under the Biden administration when CAISI was known as the U.S. Artificial Intelligence Safety Institute. CAISI, which serves as the government's main hub for AI model testing, said it had already completed more than 40 evaluations, including on cutting-edge models not yet available to the public. Developers frequently hand over versions of their models with safety guardrails ⁠stripped back so the center can probe for national security risks, the agency said. Last week, the Pentagon said it had reached agreements with seven AI companies to deploy their advanced capabilities on the Defense Department's classified networks as it seeks to broaden the range of AI providers working across the military. Microsoft, Google and xAI did not immediately respond to requests for comment.

Factbox-What We Know About US Stress Tests of Google, XAI and Microsoft AI Models

By Courtney Rozen and Jody Godoy WASHINGTON, May 5 (Reuters) - The Trump administration ⁠on ⁠Tuesday announced it had expanded a program ⁠to give U.S. government scientists access to unreleased artificial intelligence models to conduct risk assessments to include Google's DeepMind, xAI and Microsoft. ChatGPT maker OpenAI and Claude owner Anthropic had already been voluntarily working with the U.S. Center for AI Standards and Innovation, the team of U.S. government scientists, to test unreleased models for vulnerabilities, according to the companies. Here is what we know about the ⁠reviews: WHAT RISKS ARE ⁠THE U.S. FOCUSED ON? U.S. government scientists are focused on "demonstrable risks," such as the risk that advanced models can be used to launch cyberattacks on American infrastructure, according to the CAISI website. They want to limit opportunities for U.S. adversaries to use AI to develop chemical or biological weapons, or corrupt the data used to train American AI models. WHAT WILL COMPANIES HAND OVER? OpenAI is working with the group to test GPT-5.5-Cyber, said Chris Lehane, head of global affairs at OpenAI, in ⁠a ⁠LinkedIn post on Tuesday. GPT-5.5-Cyber is ⁠a variant of its latest model designed for defensive cybersecurity work. Microsoft will work with the scientists to build shared datasets and workflows to assess advanced AI models, the ⁠company said in a statement. Microsoft did not specify which models. Anthropic gave CAISI access to both publicly available and unreleased models, allowing researchers to probe for vulnerabilities in a process known as "red-teaming," or simulating the behavior of malicious actors, the company said in September. The company also gave CAISI detailed documentation on known vulnerabilities and safety mechanisms. Google DeepMind, Alphabet's AI research arm, will provide access ⁠to its "proprietary models" and data, a spokesperson said. xAI did not immediately respond to a request for comment ⁠from Reuters. WHAT HAS THE U.S. FOUND SO FAR? Anthropic's work with CAISI revealed that tricks such as claiming that human review had occurred, or substituting characters, could get around safety mechanisms, the company said, adding that it had patched the vulnerabilities. OpenAI said in September that it worked with CAISI to probe vulnerabilities in its ChatGPT Agent that could have allowed sophisticated actors to bypass OpenAI's cybersecurity measures. The exploit would have allowed the attacker to "remotely control the computer systems the agent could access for that session and successfully impersonate the user for other websites they'd logged into," the company said. The companies, along with Meta, Amazon and Inflection AI, agreed in 2023 to ⁠allow independent experts to check their models for biosecurity and cybersecurity risks. The U.S. government scientists, organized under a different name during former U.S. President Joe Biden's tenure, also released voluntary guidelines to protect against the risk of AI models leaking private health information or producing incorrect answers. The scientists are now working on guidelines for critical infrastructure providers, such as the communications and emergency services sectors, to test their AI systems, according to their website. (Reporting by Courtney Rozen; Editing by Stephen Coates)

Microsoft, Google And xAI Collaborate With US Government On Early AI Model Risk Evaluations - Amazon.com

The Center for AI Standards and Innovation (CAISI) under the Department of Commerce announced the agreement on Tuesday. The deal would allow the center to examine and research the AI models to determine their capabilities and potential security risks before their public deployment. The body has already conducted over 40 evaluations, including on advanced models that are not yet publicly available. Developers often submit versions of their models with safety guardrails removed, enabling the center to thoroughly assess national security risks. Chris Fall, the Director of CAISI, underscored the necessity of independent and rigorous measurement science in comprehending the impact of frontier AI on national security. This agreement builds on the 2024 deals with OpenAI and Anthropic, which were established during the Biden administration when CAISI was known as the U.S. Artificial Intelligence Safety Institute. Furthermore, Alphabet's president of global affairs, Kent Walker, confirmed Google's commitment to supporting defense agencies with AI. Walker assured that Google's support for military uses of AI is in line with the approaches of other major AI labs. Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors. Image via Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.

White House mulls order directing government executives to review AI models before public review. What does it mean for Aritificial Intelligence?

Donald Trump had positioned himself as a champion of unfettered AI development, rolling back Biden-era safety evaluation requirements and casting regulation as a threat to US competitiveness with China. President Donald Trump's administration is considering requiring US government oversight of artificial intelligence models before they are released to the public, a sharp reversal of the previous hands-off approach to the technology, The New York Times reported Monday. The White House is discussing an executive order that would establish a working group of tech executives and government officials to examine potential review procedures for new AI models, the newspaper reported, citing US officials and people with knowledge of the deliberations. Senior administration officials briefed executives from Anthropic, Google and OpenAI on some of the plans in meetings last week, according to the report. The shift marks a dramatic pivot from the administration's earlier stance. Trump had positioned himself as a champion of unfettered AI development, rolling back Biden-era safety evaluation requirements and casting regulation as a threat to US competitiveness with China. The Biden administration had issued an executive order in 2023 that required AI developers to share safety test results with the government and directed federal agencies to set standards for the technology, measures Trump rescinded shortly after taking office. But mounting public anxiety over AI's impact on jobs, energy costs, education and mental health -- along with bipartisan concern in Congress -- appears to have shifted the calculus, the Times reported. The immediate catalyst was the emergence of a powerful new AI model called Mythos, built by the San Francisco start-up Anthropic. The company has described the model's ability to identify software security vulnerabilities as potentially leading to a cybersecurity reckoning and has declined to release it publicly. Administration officials want to avoid political fallout from a devastating AI-enabled cyberattack and are also evaluating whether advanced models could yield capabilities useful to the Pentagon and intelligence agencies, the report said. A White House official told the Times that buzz about a potential executive order was "speculation" and said Trump would make any policy announcement himself. The rethink comes amid a leadership shake-up on AI within the Trump administration. Silicon Valley insider David Sacks, the former White House AI czar who championed deregulation, departed from the role in March. White House Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent have stepped in, telling associates they intend to take a more active hand in shaping AI policy. Their efforts have been complicated by an ongoing legal battle between Anthropic and the Pentagon over a $200 million contract and the terms of AI use by the US military.

Tech Giants Open Frontier AI Models to Federal National Security Tests | PYMNTS.com

They join Anthropic and OpenAI, which signed agreements with CAISI's predecessor, the U.S. Artificial Intelligence Safety Institute, in August 2024 and later renegotiated those agreements to match CAISI's directives, according to the release. CAISI is the government's primary point of contact for the AI industry, and it works with the industry on testing, collaborative research and the development of best practices. The agency has completed more than 40 pre-deployment evaluations of AI models, including some models that have not been released. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," CAISI Director Chris Fall said in the release. "These expanded industry collaborations help us scale our work in the public interest at a critical moment." This announcement follows several reports of concerns about artificial intelligence posing a threat to cybersecurity. Treasury Secretary Scott Bessent told Fox News Sunday (May 3) that when he and Federal Reserve Chair Jerome Powell met with Wall Street executives on concerns about cyberthreats related to Anthropic's Mythos AI model, they told the banks they should take the model seriously and use it to find holes in their defenses. "What we've had in the past month was a step change in the power of one large language model, but we're going to see it from the other AI companies, and it's important that the U.S. stays ahead here," Bessent said. It was reported April 7 that Anthropic was allowing select partners early access to Claude Mythos Preview, a model positioned for defensive cybersecurity work, so that they could identify vulnerabilities and strengthen systems before threats could be exploited. The World Economic Forum's Global Cybersecurity Outlook 2026, released in January, found that AI is expected to be the most consequential factor shaping cybersecurity strategies this year, with 94% of surveyed executives citing the technology as a force multiplier for both defense and offense.

Microsoft, Google, xAI agree to share AI models with White House for security reviews

Microsoft, Google's DeepMind and Elon Musk's xAI have agreed to share early versions of their powerful AI models with the US government for pre-clearances and security reviews, the Department of Commerce said Tuesday. The department's Center for AI Standards and Innovation said it will "conduct pre-deployment evaluations and targeted research" to better understand the capabilities and risks that come with new tools. Previous agreements with Anthropic and OpenAI have also "been renegotiated" to reflect Commerce Secretary Howard Lutnick and President Trump's new directives on security reviews, the center said. The moves come amid mounting fears over new AI tools like Anthropic's Mythos, which the company's execs warned could cause a wave of hacks and terror attacks if it ever fell into the wrong hands. Google DeepMind declined to comment. Tom Lue, the research lab's vice president of global AI affairs, confirmed the partnership in a social media post Tuesday. The White House, Department of Commerce, Microsoft and xAI did not immediately respond to The Post's requests for comment. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," Chris Fall, director of the Center for AI Standards and Innovation, or CAISI, said in a statement Tuesday. "These expanded industry collaborations help us scale our work in the public interest at a critical moment," he added. Fall was recently announced as the center's director after ex-Anthropic researcher Collin Burns was pushed out following just four days on the job, according to the Washington Post. The change in leadership came as the White House and Anthropic have been feuding over safety policies. The AI Safety Institute was established in 2023 under the Biden administration. It was renamed as CAISI under the Trump administration, with the White House's aiming to lift AI safety guardrails to boost the rollout of new models. Trump previously touted the need for rapid tech acceleration with the goal of beating China in the global AI race. But Anthropic's controversial rollout of Mythos has given some policymakers pause. A nightmarish analysis from Anthropic itself showed that Mythos could easily exploit electric grids, power plants and hospitals if hacked. The model has already "found thousands of high-severity vulnerabilities, including some in every major operating system and web browser," the AI company previously trumpeted. It's said access would be limited to a group of companies including Amazon, Google and JPMorgan. Anthropic CEO Dario Amodei has predicted that other rivals will catch up in months. OpenAI is planning a limited release of its latest model, called GPT-5.5-Cyber, over security concerns. Americans appear to be growing skeptical of AI innovation. A Pew Research Center poll last year found that 50% of Republicans and 51% of Democrats said they were more concerned than excited about the increased use of AI in daily life.

AI labs should pass safety review to get US government contracts, group says

May 11 (Reuters) - The Trump administration should screen cutting-edge artificial intelligence models for security threats before they are publicly released and withhold lucrative government contracts from those that fail review, an advocacy group told U.S. officials on Monday. The White House is grappling with the implications of Anthropic's Mythos, which could make complex cyberattacks easier and quicker to execute, posing national security risks. Americans for Responsible Innovation urged the Trump administration to develop methods to vet upcoming frontier models from larger developers for cyberattack and weapons development capabilities. Companies should have to pass the review to be eligible for government contracts, the group said in a letter to administration officials. The U.S. Center ?for AI Standards and Innovation already reviews some AI models through voluntary agreements with OpenAI, Anthropic, and, more recently, Google, Microsoft and xAI. CAISI should take the lead on developing mandatory requirements, and Congress should create a permanent enforcement office within the U.S. Department of Commerce to enforce the requirements, the group said. The proposed requirements would apply to companies that spend $100 million or more a year on compute to train frontier models, or that make at least $500 million in revenue annually from AI products and services. California has a similar threshold for safety reporting requirements enacted last year. (Reporting by Jody Godoy in New YorkEditing by Rod Nickel)

What we know about US stress tests of Google, xAI and Microsoft AI models

WASHINGTON, May 5 (Reuters) - The Trump administration on Tuesday announced it had expanded a program to give U.S. government scientists access to unreleased artificial intelligence models to conduct risk assessments to include Google's DeepMind, xAI and Microsoft. ChatGPT maker OpenAI and Claude owner Anthropic had already been voluntarily working with the U.S. Center for AI Standards and Innovation, the team of U.S. government scientists, to test unreleased models for vulnerabilities, according to the companies. U.S. government scientists are focused on "demonstrable risks," such as the risk that advanced models can be used to launch cyberattacks on American infrastructure, according to the CAISI website. They want to limit opportunities for U.S. adversaries to use AI to develop chemical or biological weapons, or corrupt the data used to train American AI models. WHAT WILL COMPANIES HAND OVER? OpenAI is working with the group to test GPT-5.5-Cyber, said Chris Lehane, head of global affairs at OpenAI, in a LinkedIn post on Tuesday. GPT-5.5-Cyber is a variant of its latest model designed for defensive cybersecurity work. Microsoft will work with the scientists to build shared datasets and workflows to assess advanced AI models, the company said in a statement. Microsoft did not specify which models. Anthropic gave CAISI access to both publicly available and unreleased models, allowing researchers to probe for vulnerabilities in a process known as "red-teaming," or simulating the behavior of malicious actors, the company said in September. The company also gave CAISI detailed documentation on known vulnerabilities and safety mechanisms. Google DeepMind, Alphabet's AI research arm, will provide access to its "proprietary models" and data, a spokesperson said. xAI did not immediately respond to a request for comment from Reuters. WHAT HAS THE U.S. FOUND SO FAR? Anthropic's work with CAISI revealed that tricks such as claiming that human review had occurred, or substituting characters, could get around safety mechanisms, the company said, adding that it had patched the vulnerabilities. OpenAI said in September that it worked with CAISI to probe vulnerabilities in its ChatGPT Agent that could have allowed sophisticated actors to bypass OpenAI's cybersecurity measures. The exploit would have allowed the attacker to "remotely control the computer systems the agent could access for that session and successfully impersonate the user for other websites they'd logged into," the company said. The companies, along with Meta, Amazon and Inflection AI, agreed in 2023 to allow independent experts to check their models for biosecurity and cybersecurity risks. The U.S. government scientists, organized under a different name during former U.S. President Joe Biden's tenure, also released voluntary guidelines to protect against the risk of AI models leaking private health information or producing incorrect answers. The scientists are now working on guidelines for critical infrastructure providers, such as the communications and emergency services sectors, to test their AI systems, according to their website. (Reporting by Courtney Rozen; Editing by Stephen Coates)

Microsoft, Google and xAI to give US government early access to AI models for security checks

May 5 (Reuters) - Microsoft, Alphabet-owned Google and Elon Musk's xAI will give the U.S. government early access to new artificial intelligence models before their public release to allow checks for national security risks under a new deal. The Center for AI Standards and Innovation at the Department of Commerce said on Tuesday that the agreement would allow it to evaluate the models before deployment and conduct research to assess their capabilities and security risks. The development of advanced AI systems including Anthropic's Mythos has in recent weeks created a stir globally, including among U.S. officials and corporate America, over their ability to supercharge hackers. "Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications," CAISI Director Chris Fall said in a statement. The move builds on agreements with OpenAI and Anthropic, established in 2024 under the Biden administration when CAISI was known as the U.S. Artificial Intelligence Safety Institute. CAISI, which serves as the government's main hub for AI model testing, said it had already completed more than 40 evaluations, including on cutting-edge models not yet available to the public. Developers frequently hand over versions of their models with safety guardrails stripped back so the center can probe for national security risks, the agency said. Microsoft and xAI did not immediately respond to requests for comment. Google declined to comment. Last week, the Pentagon said it had reached agreements with seven AI companies to deploy their advanced ?capabilities on the Defense Department's classified networks as it seeks to broaden the range of AI providers working across the military. The Pentagon announcement did not include Anthropic, which has been embroiled in a dispute with the Pentagon over guardrails on the military's use of its AI tools. (Reporting by Aditya Soni in Bengaluru, additional reporting by Jaspreet Singh in Bengaluru; Editing by Shinjini Ganguli and Mrigank Dhaniwala)

White House considers government reviews for AI models, NYT reports

WASHINGTON, May 4 (Reuters) - U.S. President Donald Trump is considering the introduction of government oversight over new models of artificial intelligence, the New York Times reported on Monday, citing officials briefed on the deliberations. The U.S. government is discussing an executive order to create an AI working group that would bring together tech executives and government officials to examine potential oversight procedures, according to the newspaper. A White House official declined to confirm or deny the report. "Any policy announcement will come directly from the president. Discussion about potential executive orders is speculation." The newspaper said the White House was considering a formal government review process for new AI models. The Times said the change could be prompted by concerns about Anthropic's new AI model called Mythos, which cybersecurity experts warn could supercharge complex cyberattacks. Its capabilities to code at a high level have given it a potentially unprecedented ability to identify cybersecurity vulnerabilities and devise ways to exploit them, experts said. The move would be a sharp reversal for Trump, who has urged a hands-off approach. Trump in July released an AI blueprint that aimed to loosen environmental rules and vastly expand AI exports to allies, in a bid to maintain the American edge over China in the critical technology. On Trump's first day in office in 2025, he revoked a 2023 executive order signed by Joe Biden that sought to reduce the risks that artificial intelligence poses to consumers, workers and national security. Biden's order required developers of AI systems that pose risks to U.S. national security, the economy, public health or safety to share the results of safety tests with the U.S. government, in line with the Defense Production Act, before they were released to the public. The White House in March unveiled an AI policy for Congress that urges lawmakers to enact legislation to pre-empt state rules, protect children ?and shield communities from high energy costs related to the burgeoning technology. (Reporting by David Shepardson and Alex Alper in Washington and Chandni Shah in Bengaluru; Editing by Sonali Paul)