In a move that can only be described as 'launch, then immediately regret it,' Anthropic has completely shut off access to its Mythos 5 and Fable 5 models, just days after they burst onto the scene with all the fanfare of a confetti cannon at a funeral.

Friday evening, the US Commerce Department handed Anthropic a directive, essentially putting the new models under export controls that restrict their use to within the United States. Anthropic, in a Friday night message, explained that the only way to comply with the government's order 'is that we must abruptly disable Fable 5 and Mythos 5 for all our customers.' Other Anthropic models remain unaffected, so ChatGPT users can rest easy.

An Axios report, citing an administration official, says the administration is twitchy about reports of a jailbreak that allegedly bypasses the broad classifier-based safeguards designed to keep Fable 5 from helping with cybersecurity, chemistry, and biology prompts. The administration reportedly asked for a pause to give the 'national security apparatus' time to 'harden' against this threat - a process that could take 'the next few weeks,' according to the source. Because nothing says 'national security' like panicking over a hypothetical jailbreak.

Anthropic, in its Friday night announcement, noted that the government has only provided 'verbal evidence of a potential narrow, non-universal jailbreak' involving Fable 5 reviewing a specific codebase for software flaws. The company says it has only seen this jailbreak used to find 'minor' and 'relatively simple' vulnerabilities, and that other publicly available models like GPT-5.5 can do similar things. But apparently, when you're the new kid on the block, you get held to a different standard.

'We are complying with the government's legal directive and are removing access to Fable 5 and Mythos 5 for all users,' Anthropic writes, in the tone of a kid who's been told to go to their room. 'However, 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. If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.'

Earlier this month, President Trump signed an executive order encouraging AI model makers to submit to voluntary government security testing. That order came after a planned signing ceremony was abruptly postponed last month amid reports of internal disagreements. Because what's a government directive without a little chaos?

Anthropic apologized to customers for the 'disruption,' which it says is the result of a 'misunderstanding,' and promised more details within 24 hours. So stay tuned for the next episode of 'AI Drama: Government Edition.'