White House Nears Frontier AI Deal With OpenAI, Anthropic and Google — Announcement Due by August 1

The White House is expected to announce a voluntary frontier AI framework with OpenAI, Anthropic and Google before August 1, when the 60-day deadline set by President Trump's June 2 executive order on AI and cybersecurity expires.

The deal in brief

• 30-day window: Participating labs would give federal agencies advance access to "covered frontier models" for up to 30 days before public release, for security evaluation.
• Classified benchmarks: The tests used to evaluate the models are classified, developed by Treasury, Defense and Homeland Security under the executive order.
• Voluntary, for now: The framework carries no statutory force — it formalizes the pre-deployment testing relationships labs already had with the US AI Safety Institute, but with sharper teeth and clearer timelines.
• Meta is out: The most notable absence. Meta's open-weight release strategy sits awkwardly with a pre-release government review window, and the company has not joined.

Why it matters

The framework would be the Trump administration's most concrete frontier-governance move to date, arriving amid its parallel deliberations over banning Chinese open-weight models — a debate sharpened by Kimi K3's launch and this week's letter from nearly 200 startups opposing restrictions.

For Asia's labs and governments, the structure is worth studying: it establishes pre-release state access as the emerging US norm for frontier systems. Seoul, Tokyo and Singapore — each drafting their own frontier-model rules — now have a template, and Chinese labs releasing open weights have a counter-model. Expect the announcement in the first week of August at the latest.