Trump's AI Executive Order Hits Its August 1 Deadline, Formalizing a 30-Day Review Window for Frontier AI Models

My take: Today marks the 60-day deadline President Trump set in his June executive order for the government to formalize how it reviews the most advanced AI models before they reach the public. Until now, interventions were improvised: Claude Fable 5 was suspended for three weeks and GPT-5.6 was restricted for 12 days, both without a published process or clear rules. That changes today.

The new voluntary framework defines a window of up to 30 days in which the federal government can access a frontier model before launch, run its own cybersecurity capability assessments, and determine which trusted partners can receive it. OpenAI and Anthropic have already agreed to cooperate; the question is whether Meta and xAI will do the same once their models cross the defined capability threshold.

For those who use these tools at work, this matters: the process for how a new AI model reaches the market is taking official shape. A transparent process with published rules is better than ad hoc decisions no one can anticipate or prepare for.

The question that remains open is whether this framework will have enough reach to include those who prefer to operate without rules, or whether it will end up as a courtesy the large labs follow while everyone else ignores it.