US pulls Anthropic's two most powerful AI models
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic received the shutdown directive Friday at 5:21 pm ET and announced on X that it has complied, disabling Claude Fable 5 and Claude Mythos 5 for all users globally — not just the foreign nationals the export control nominally targeted.
- Claude Mythos 5, Anthropic's most capable model previewed in early April, identified security flaws in every major operating system and web browser it tested, prompting Anthropic to restrict it to roughly 50 vetted organizations via Project Glasswing, including Amazon, Apple, Google, Microsoft, and CrowdStrike.
- Claude Fable 5, released just three days before the shutdown, was the guardrailed commercial version of Mythos and was immediately the most capable AI model available to the public according to Vals AI benchmark tests.
- Anthropic says the government's actual concern is a claimed 'potential narrow, non-universal jailbreak' of Fable 5 — prompting the model to read a specific codebase and flag software flaws — for which the government has provided only verbal evidence.
- Anthropic argued the jailbroken capability is already widely available in other public models, including OpenAI's GPT-5.5, and that its independent classifier systems provide additional safeguards beyond the model itself.
- Anthropic publicly pushed back: 'We disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people,' warning the standard 'would essentially halt all new model deployments for all frontier model providers.'
- OpenAI's Sam Altman had predicted this dynamic in April, calling Anthropic's handling of Mythos 'fear-based marketing' — 'It is clearly incredible marketing to say, we built a bomb, we were about to drop it on your head, we will sell you a bomb shelter for $100 million.'
Why it matters: Anthropic, widely expected to pursue an IPO this year, saw its flagship commercial product pulled days after launch — a direct hit to a company that built its public identity on being the safety-conscious alternative. More broadly, if the government applies this jailbreak-recall standard to other frontier model providers, Anthropic argues it would halt all new model deployments industry-wide, putting every major AI lab's release pipeline at regulatory risk.
