President Donald Trump signed an executive order creating a voluntary framework for early government access to frontier AI models. The measure follows Treasury secretary Scott Bessent’s warning about Anthropic’s upcoming Mythos model, which can locate and exploit cyber vulnerabilities. By offering a limited review path, the administration hopes to balance security concerns with industry growth.

The order mandates AI labs submit new models for a security review lasting up to 30 days before public release, a reduction from the draft’s 90‑day window. Venture‑capitalist adviser David Sacks praised the “game changer” language, saying it lets firms comply without delaying launches and applies only to models that represent a substantial cyber‑capability shift, not routine updates.

Critics within the president’s circle, including Steve Bannon and former NEC director Kevin Hassett, pushed for stricter controls, likening AI oversight to FDA drug approval. Treasury officials stress the directive is not new regulation but a safety step, while the administration plans to enlist banks and other trusted partners to vet models, laying groundwork for future licensing.