The U.S. government is holding advanced discussions with AI companies to establish voluntary standards for releasing new models, with a possible announcement as early as next week, according to the Financial Times citing sources. Washington has increased scrutiny of model releases to identify risks, amid worries that advanced AI could be misused by military or intelligence entities in China, Russia or other nations of concern. The standards would define benchmarks for advanced models and timelines, while clarifying access rules inside and outside the United States, the report said. Reuters could not immediately confirm the details. The White House, Anthropic and OpenAI did not respond to requests for comment outside business hours. In June, President Donald Trump issued an executive order directing agencies to collaborate with leading AI developers on testing advanced models before release and to prepare related standards. Google has discussed the matter with officials ahead of launching more capable coding models, a source told Reuters, noting the company is also part of wider talks on industry standards. The U.S. Commerce Department on Tuesday removed export controls on Anthropic’s top Fable and Mythos models, weeks after suspending them over national security issues. OpenAI recently postponed the full public release of GPT-5.6 at the government’s request, restricting access to a limited group of approved partners. Both companies are preparing for IPOs.
