Just days before the G7 summit in France, the Trump administration barred export of Anthropic's most powerful AI models — Fable 5 and Mythos 5 — a move that dominated discussions among world leaders and left the company's CEO, Dario Amodei, pressing for international cooperation. Amodei urged attendees on June 17 not to let the dispute fracture global AI governance, but he left France without a resolution.
A Ban Before the Summit
The US Commerce Department issued a directive to Anthropic, blocking Fable 5 and Mythos 5 from being used by non-US customers and even foreign nationals inside the United States. Commerce Secretary Howard Lutnick pointed to a reported jailbreak of Mythos 5 that, he said, would let users bypass safety guardrails to extract software vulnerability data. The order came days before the G7 meeting in France, where AI policy was a central topic.
This wasn't the first US action against Anthropic. In February, Trump ordered all federal agencies to stop using Anthropic's products after the company refused Pentagon contract terms that would have required its AI to be available for 'any lawful purpose'.
The Jailbreak That Triggered It
Lutnick specifically cited the Mythos 5 jailbreak as the reason for the export ban. According to the Commerce Department, the vulnerability would allow malicious actors to bypass safety filters and pull out sensitive code or security flaws. Anthropic has not confirmed the jailbreak's existence publicly, but the company's senior staff have been sent to Washington to seek a reversal of the ban.
Anthropic warned that applying the same standard across the industry would effectively halt all new model deployments. The company argues that any frontier AI can theoretically be jailbroken, and a ban based on that risk would freeze innovation.
Allies Push Back
French President Emmanuel Macron said the dispute clarified the stakes. If the US can 'turn off the switch' on AI models, he argued, it would damage American companies that are currently leading the global AI race. Indian Prime Minister Narendra Modi raised a different concern: democratic nations must retain access to frontier AI to protect critical infrastructure.
OpenAI's Sam Altman and Google DeepMind's Demis Hassabis backed Amodei during the G7 discussions, sitting at the same table. Their support signals that even competitors see the ban as a threat to the broader AI ecosystem.
Anthropic's Washington team is pushing for the Commerce Department to lift the restrictions. The company argues the ban sets a dangerous precedent — one that could be applied to any model with a reported flaw. The G7 ended without a joint statement on AI export controls, and Amodei left France without a deal. For now, Fable 5 and Mythos 5 remain locked to US soil, and the question of whether other Allied nations will develop their own restrictions hangs unresolved.
