U.S. government in advanced talks with AI firms on voluntary model release standards, FT reports
July 1 (Reuters) – The U.S. government is in advanced talks with major artificial intelligence companies to craft voluntary standards governing the release of next-generation AI models, with an announcement potentially coming as soon as next week, the Financial Times reported Wednesday, citing people familiar with the discussions.
Washington has ramped up scrutiny of new model deployments to flag potential risks, amid growing worries that advanced AI systems could be exploited by military intelligence agencies in China, Russia or other adversarial nations. The proposed voluntary framework would establish benchmarks for cutting-edge models and define timelines for when they can be made public, while also clarifying which domestic and international users may gain access, according to the report.
The White House, Anthropic and OpenAI did not immediately respond to Reuters’ requests for comment outside regular business hours. Reuters could not independently verify the FT’s reporting.
The negotiations build on a June executive order in which U.S. President Donald Trump directed federal agencies to collaborate with leading AI developers on pre-release testing of advanced models and to draft safety standards. The order marked one of the administration’s most aggressive moves to assert oversight over a rapidly evolving technology sector.
Google has been in direct talks with the government ahead of releasing advanced coding models endowed with more sophisticated capabilities, a source told Reuters on Wednesday. That source added that Google was also participating in broader industry-wide discussions on model standards. The Financial Times first reported those details.
In a related development, the U.S. Commerce Department on Tuesday lifted export restrictions on Anthropic’s most advanced models, dubbed Fable and Mythos, less than three weeks after those restrictions were imposed over national security concerns. The reversal underscores the delicate balance the administration is trying to strike between fostering innovation and guarding against potential misuse.
OpenAI has also faced government constraints. Last week, the company delayed the full public launch of GPT-5.6 at the request of U.S. officials, limiting initial access to a small group of pre-vetted partners. Both OpenAI and Anthropic are preparing for initial public offerings, adding financial stakes to the regulatory maneuvering.
Industry analysts say the voluntary standards approach, if finalized, could serve as a template for other nations grappling with AI governance, but also note that absent binding rules, compliance may remain uneven. The talks come as lawmakers in Congress continue to debate more comprehensive AI legislation.
(Reporting by Carlos Méndez in Mexico City and Preetika Parashuraman in Bengaluru; Additional reporting by Courtney Rozen and Mrinmay Dey; Editing by Maju Samuel and Rashmi Aich)