The government told Anthropic that it became aware of a method to bypass or jailbreak Fable 5. Anthropic reviewed the technique and said what it saw was narrow, not a universal jailbreak, and involved the identification of a small number of previously known minor vulnerabilities. He said other publicly available models, including OpenAI’s GPT-5.5, can detect the same vulnerabilities without any workarounds.
The company said the government has so far provided only verbal evidence of a possible narrow jailbreak, which it described as essentially asking the model to read a code base and fix software flaws, a task defenders use daily.
He said enforcing this standard across the industry would “essentially stop all new model deployments for all frontier model providers.”
Anthropic has built its entire brand around the development of security-focused AI and is now publicly challenging a national security directive on the grounds that the government’s evidence does not cross its own boundaries.
The company will share more details about the specific jailbreak within 24 hours.
The crypto market now views the shutdown as a negative for the IPO case, and Anthropic’s decline from its post-launch highs reflects that. The first question regarding the company’s public listing ambitions is whether the government order will be rescinded, restricted or extended to other model classes once Anthropic issues its technical rebuttal.
