Claude Fable 5 was banned for 18 days – what happened (freemalta.com)

Anthropic's Claude Fable 5 was temporarily banned for 18 days after the US Department of Commerce raised export controls following concerns about a jailbreak technique discovered by Amazon researchers. The technique allowed Fable 5 to identify software vulnerabilities, prompting the government to classify it as a national security issue. To comply with the directive, Anthropic swiftly pulled Fable 5 offline globally, including for its non-US employees, while it worked on a solution. After negotiations and the implementation of an enhanced safety classifier that effectively blocks harmful requests, Fable 5 was restored on July 1. This episode is significant for the AI/ML community as it underscores the complexities of governmental oversight and the potential of security implications associated with advanced models. While Claude Fable 5 and its more powerful counterpart, Mythos 5, share the same architecture, Fable 5 is equipped with stricter safety measures to prevent misuse in cybersecurity tasks, making it more suitable for broader commercial applications. Developers must now navigate new constraints and pricing changes, as the refined safety classifier is expected to generate more false positives in security-related inquiries, affecting production workflows that rely on AI for coding and debugging tasks.