Anthropic's Claude Fable 5 is a new AI model based on the same architecture as the Mythos model, which was identified as having potential security risks due to its capability to assist attackers in finding and exploiting software vulnerabilities. To address these risks, Fable 5 incorporates strict safeguards that block or divert sensitive queries related to offensive cybersecurity, biology, or chemistry to a previous, less powerful model (Opus 4.8). The unrestricted Mythos 5 model remains accessible only to a highly vetted group of trusted partners, including government cyberdefenders and select researchers. Fable 5 is offered free for a limited time to certain customers but consumes tokens at a significantly higher rate, limiting its widespread availability. Anthropic's approach reflects a cautious rollout to prevent misuse while enabling defenders to leverage advanced AI capabilities.

The underlying Mythos model's capabilities could potentially be exploited by attackers to identify and exploit vulnerabilities in software, posing a security risk to organizations. However, the Fable 5 model includes strong guardrails to prevent misuse by blocking sensitive queries, reducing the risk of exploitation. Access to the unrestricted Mythos model is tightly controlled, limiting exposure. The high token consumption of Fable 5 may also limit its use in large-scale automated attacks. There are no known exploits in the wild at this time.

Anthropic has implemented strict safeguards in the Fable 5 model to block sensitive queries and limit misuse. The unrestricted Mythos model is only available to vetted trusted partners, reducing risk. Organizations should use the Fable 5 model within the provided guardrails and monitor any access to the Mythos model to ensure it is limited to authorized users. Since this is a cloud service, Anthropic manages the security and remediation of the models. Patch status is not applicable. No additional immediate action is required beyond following Anthropic's access controls and usage policies.