US in talks with tech companies on voluntary standards for new AI models

WASHINGTON — The U.S. government is in advanced talks with leading artificial intelligence companies to establish voluntary standards governing the release of new AI models, with an announcement possible as soon as next week, the Financial Times reported on Wednesday, citing sources.

The proposed framework reflects Washington’s growing effort to tighten oversight of advanced model deployments amid concerns that powerful AI systems could be misused by military or intelligence actors in countries including China, Russia, and other nations of concern.

Under the draft approach, the standards would set performance benchmarks for frontier AI models and establish timelines for safety testing prior to release. They would also clarify rules on who can access advanced systems, both within the United States and abroad, according to the report.

In June, U.S. President Donald Trump signed an executive order directing federal agencies to work closely with leading AI developers to evaluate advanced models before they are released publicly and to develop formal safety standards.

The move comes as U.S. regulators and technology firms increasingly navigate the balance between innovation and national security. On Tuesday, the U.S. Commerce Department lifted export controls on Anthropic’s most advanced Fable and Mythos models, less than three weeks after temporarily suspending their release over security concerns.

OpenAI has also faced regulatory constraints. Last week, the company delayed a full public rollout of GPT-5.6 at the request of the U.S. government, limiting access to a small group of vetted partners. Both OpenAI and Anthropic are reportedly preparing for initial public offerings.

Google has also been in discussions with the U.S. government ahead of the release of its next-generation coding models, which are expected to feature more advanced cybersecurity capabilities than previous versions, the Financial Times reported, citing sources.

Reuters said it was unable to independently verify the report. The White House, Anthropic, and OpenAI did not immediately respond to requests for comment outside of regular business hours. Google declined to comment.

-Reporting by Carlos Méndez in Mexico City, additional reporting by Courtney Rozen and Mrinmay Dey; Editing by Maju Samuel and Rashmi Aich/Reuters