Anthropic has taken its Claude Fable 5 AI model offline in compliance with a US government order, according to WIRED AI. In a blog post, the company stated that “the government believes it has become aware of a method of bypassing, or ‘jailbreaking’ Fable 5.”
The action represents a rare instance of government intervention in AI model deployment based on security concerns. Jailbreaking refers to techniques that bypass an AI system’s safety guardrails, potentially allowing it to generate harmful or prohibited content that its developers intended to prevent.
Anthropic did not provide additional details about the specific jailbreaking method that prompted the government’s concern or the timeline for when Fable 5 might return to service. The company’s decision to comply with the order highlights the growing scrutiny AI systems face from regulatory authorities regarding potential misuse and safety vulnerabilities.