Kevin Hassett, Director of the National Economic Council, has confirmed that the White House is working on an executive order that would require AI models to undergo a vetting process before release. The template? FDA drug approval.
What’s Being Planned
The details aren’t final yet, but the direction is clear: before an AI model with certain capabilities hits the market, it would need to be tested and approved. Hassett said the process would ‘very likely’ apply to all AI companies — not just Anthropic.
The Commerce Department has already expanded a voluntary testing program that now includes Google, Microsoft, xAI, OpenAI, and Anthropic.
Why Now?
The trigger has a name: Mythos. Anthropic’s most powerful model demonstrated what happens when an AI systematically discovers software vulnerabilities. Mozilla ran Mythos against Firefox and found 271 security flaws in a single evaluation pass — more than human teams had found in 18 months.
That set off alarm bells in Washington. If an AI model can find hundreds of zero-days in hours, the question becomes: what happens when that model falls into the wrong hands?
A Course Correction for Trump
What makes this politically interesting is the context. The Trump administration has consistently pushed deregulation. AI was supposed to flourish, not be held back. Biden’s AI executive order was revoked. And now? Now the same administration is proposing FDA-style pre-release testing.
The difference: this isn’t about bias or fairness — it’s about national security. And when it comes to security, even this administration doesn’t hold back.
What This Means for the Industry
If this order goes through, it would be the first time the US introduces mandatory pre-release testing for AI models. That would have massive implications for release cycles, open-source models, and international competitiveness.
The key question: will this only apply to models above a certain capability threshold? Or to everything? The details will determine whether this becomes a smart safeguard — or regulatory dead weight.
Sources: Bloomberg · The Hill · Yahoo Finance