Signal US in advanced talks with AI companies over voluntary standards for releasing powerful models

Summary

As of early July 2026, the US government was in advanced talks with major AI companies over voluntary standards for releasing powerful models, covering benchmarks, release schedules and access rules, with the Financial Times reporting an announcement possible as soon as the following week. The talks trace back to a June 2 executive order from President Trump directing Treasury, Defense and Homeland Security to develop a benchmarking process and voluntary framework for advanced AI models within 60 days, with that deadline expiring in the first week of August. By early July, the White House was reportedly nearing a voluntary agreement specifically with OpenAI, Anthropic and Google — but not Meta — that would give federal agencies a 30-day pre-release review window on frontier models, though the benchmarks used remain classified. As of July 3, GPT-5.6 variants (Sol, Terra and Luna) remained limited to roughly 20 government-vetted partner organizations, with OpenAI saying it would keep coordinating with government partners while not viewing such access processes as a long-term default. The oversight push was driven partly by concerns that advanced AI could be misused by military or intelligence services in China, Russia or other adversary states. On the same day this story broke, surging AI-driven memory chip demand was separately reported to be pushing up smartphone and PC prices.