Anthropic launches Claude Sonnet 5 and restores Fable and Mythos
Claude Sonnet 5 returns as Anthropic restores frontier models after review See what changed for enterprise AI, safety controls, and software workflows
Anthropic has released Claude Sonnet 5 and restored access to its Fable and Mythos frontier models after a federal export control review.
The move ends an 18day pause that began after a U.S. government directive on June 12 forced the temporary suspension of the company’s highestcapability systems. The restriction followed a reported bypass of safety controls in Fable 5, which allowed the model to identify software vulnerabilities and generate exploitation code.
Anthropic said it updated an automated safety classifier to address the issue, and that the change cleared the way for wider commercial deployment across its platform, cloud infrastructure, and partner networks. The company also said testing showed similar vulnerabilityidentification behavior in older models from several providers.
The article also highlights Claude Sonnet 5’s performance in agentic workflows and software development tasks, including code review, debugging, and multistep administrative work. Anthropic says the model is being positioned for enterprise use, while commercial versions continue to ship with safety classifiers intended to reduce risky requests.