White House Builds Ad Hoc AI Licensing Regime
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- The Trump administration sent an export control directive to Anthropic nearly a week ago, forcing the AI lab to pull its most advanced models — Claude Mythos and Fable 5 — offline, though the US government has never publicly stated what Anthropic did wrong.
- Anthropic says it coordinated with the US government on the Mythos rollout and revoked SK Telecom's access immediately when concerns about the South Korean telecom giant's alleged China ties were raised; Anthropic had worked with the company for years without prior national security issues.
- The White House's demand that Anthropic block all foreign nationals from accessing Mythos and Fable 5 locked out Anthropic's own researchers and customers including Apple, Meta, and much of the Fortune 500, hampering the very innovation the administration claims to protect.
- Amazon CEO Andy Jassy separately raised concerns to Treasury Secretary Scott Bessent that guardrails on Claude Fable 5 could be circumvented via jailbreaking — a problem AI researchers say is inherent to probabilistic models and cannot be fully solved.
- AI lab leaders at companies including OpenAI, Google, and Meta are concluding they need to give the White House early access to forthcoming models and proactively share launch details, according to the author's interviews with executives.
- Cohere CEO Aidan Gomez said "advance notice, advance access" are the primary asks from the US and other governments, calling that engagement a positive sign for oversight of frontier technology.
- Trump's executive order created a "voluntary" model submission regime with an explicit carve-out against mandatory licensing — but the Anthropic situation has effectively turned it into an ad hoc licensing regime, according to a former White House technology official.
Why it matters: The Anthropic dispute has converted Trump's "voluntary" model submission order into a de facto mandatory licensing system with no clear written rules. Labs like OpenAI, Google, and Meta must now preemptively court the White House before each launch, and customers from Apple to much of the Fortune 500 are locked out of frontier models that Anthropic's own researchers relied on.
