Anthropic has officially launched Claude Sonnet 5 and restored access to its Fable and Mythos models following a federal export control review on June 12. This decision concludes an eighteen-day operational pause that was initiated due to concerns over the models' safety controls. The review was prompted by reports from researchers at Amazon, who discovered a method to bypass the safety measures of Fable 5.
The federal export control directive required a temporary suspension of Anthropic's highest-capability systems, leading to a total access blackout for global users. In response to these concerns, Anthropic has developed an updated automated classifier designed to patch the identified vulnerability, enabling a full commercial rollout.
Regulatory Pressures on Frontier Intelligence Systems
The operational pause of Fable 5 and Mythos 5 underscored the significant regulatory pressures that frontier intelligence systems face. During the suspension, security evaluations confirmed that the vulnerability identification behavior was not exclusive to Fable 5. Other models, including Claude Opus 4.8 and GPT-5.5, displayed similar results.
To address the federal directive, engineers trained the automated safety classifier to specifically target the bypass mechanism identified by Amazon. This classifier is designed to maintain a wide safety margin, effectively blocking ambiguous developer prompts that may indicate malicious intent.
Performance Metrics of Claude Sonnet 5
With the launch of Claude Sonnet 5, engineering teams are now shifting their focus to this model, which aims to reduce operational costs while ensuring high execution capacity. Performance data indicates that Sonnet 5 can efficiently execute multi-step plans, operate terminal environments, and navigate web browsers autonomously.
