Anthropic Redeploys Claude Fable 5 with Enhanced Safeguards

Anthropic has officially redeployed Claude Fable 5 after the US government lifted export restrictions imposed on June 12. The model, which was temporarily inaccessible due to concerns over potential misuse, is now available globally on the Claude Platform, Claude.ai, Claude Code, and Claude Cowork. For certain enterprise plans, Fable 5 will be included with usage limits through July 7, after which it will be accessible via usage credits. The company is also working to restore access on cloud platforms like AWS, Google Cloud, and Microsoft Foundry.

The initial restrictions followed a report indicating that Amazon researchers found a way to bypass Fable 5’s safeguards, revealing software vulnerabilities and demonstrating potential exploitation. In response, Anthropic collaborated with the US government and partners to develop an improved safety classifier that blocks over 99% of the bypass attempts, focusing on routine cybersecurity tasks.

Fable 5 shares the same underlying model as Mythos 5 but was launched with stronger safety measures to prevent misuse. Mythos 5, which has fewer safeguards, was initially restricted to trusted partners for cybersecurity research. The company emphasizes that Fable 5 does not possess unique offensive capabilities, and its safeguards are designed to prevent harmful misuse while allowing routine cybersecurity work.

This deployment highlights ongoing efforts to balance AI model accessibility with safety concerns, especially in sensitive areas like cybersecurity. The enhanced safeguards and collaboration with government agencies aim to mitigate risks associated with powerful AI models while enabling beneficial applications.