The partial lift applies only to Mythos 5. Anthropic's other restricted model, Fable 5, remains under export controls, though the company says it is negotiating with the government to restore general access. The specific terms governing which companies qualify as "trusted partners" have not been disclosed. OpenAI announced yesterday that it would allow the administration to screen users of its newest model, suggesting a broader shift toward government vetting of AI access across the industry.
Attorneys should monitor this as a marker of how the Trump administration intends to regulate AI exports and control access to advanced models. The rapid swing from total prohibition to selective access demonstrates the volatility of current policy and the government's willingness to impose immediate shutdowns. Companies developing or deploying advanced AI should expect increased scrutiny of their safeguards and potential restrictions tied to geopolitical concerns. The precedent of government user vetting—now accepted by both Anthropic and OpenAI—may become standard practice for future model releases.