In a surprising turn of events, Anthropic has completely halted access to its newly launched Mythos 5 and Fable 5 models just days after their debut. The decision follows a directive from the U.S. Commerce Department, which imposed export controls restricting the use of these advanced AI models outside the United States.
In a statement released Friday night, Anthropic explained that to comply with the government order, they deemed it necessary to “abruptly disable Fable 5 and Mythos 5 for all our customers.” The company reassured users that access to its other models remains unaffected.
The swift action comes in response to growing concerns within the administration regarding a potential security breach. An Axios report highlighted fears surrounding a jailbreak that could circumvent safeguards designed to block Fable 5 from engaging in discussions related to cybersecurity, chemistry, and biology. According to an administration official, this prompted a request for a pause in the release of both models, allowing time for enhancements to the national security framework to counter such vulnerabilities. Sources suggest this security strengthening could be finalized within a matter of weeks.
In the announcement, Anthropic clarified that the government provided only verbal indications of a narrow, non-universal jailbreak associated with Fable 5. This jailbreak reportedly enables the model to assess a specific codebase for software flaws. Anthropic, however, noted that the evidence of this jailbreak appears limited, as it has predominantly shown the capability to identify “minor” and “relatively simple” software vulnerabilities. The company also pointed out that other publicly available models, such as GPT-5.5, exhibit similar capabilities.
As the situation unfolds, the implications for both Anthropic and the broader AI landscape remain to be seen, particularly regarding compliance with government regulations and the ongoing ability to innovate in the field of artificial intelligence.
