U.S. Licensing Regime for Frontier AI Takes Shape as Anthropic Models Face Export Ban

Hello from Eye on AI. The U.S. licensing regime for frontier AI has quietly emerged after the Commerce Department invoked “deemed export” rules on Anthropic’s Fable and Mythos models.

U.S. licensing regime for frontier AI unveiled by Anthropic case

Amazon’s discovery of a jailbreak prompted a White House call from CEO Andy Jassy, leading to an immediate shutdown of the models for all users, including Anthropic staff. The move forces Anthropic to seek a waiver while it dispatches senior executives to Washington. “We are committed to safeguarding national security without stifling innovation,” a senior administration official told Reuters. Critics argue the action is a back‑door licensing system, repurposing existing export statutes into an opaque control framework.

Industry backlash and geopolitical ripples

Cybersecurity veterans signed an open letter warning that disabling Fable deprives defenders of a vital tool, yet the letter also noted that comparable capabilities exist in OpenAI’s GPT‑5.5, which remains unregulated.

“The policy is inconsistent and secretive, favoring entities the administration prefers,”

says Dean Ball, former AI policy adviser, in an interview with Bloomberg. Amazon has poured ↑ $13 bn into Anthropic and pledged up to ↑ $20 bn more, raising questions about how commercial stakes intersect with national security. The episode fuels European anxiety over AI sovereignty and gives Chinese open‑source developers a propaganda win. As the U.S. grapples with how to codify a transparent, fact‑based licensing process, venture capitalists may shift toward narrow‑AI startups that avoid the looming export hurdles. The future of private frontier AI development in America now hangs in the balance.