Anthropic released Claude Fable 5 on Tuesday, June 9, 2026 [3], making its most powerful AI capability available to the general public.
The release marks a significant step in AI accessibility, balancing the deployment of high-tier intelligence with strict controls to prevent the technology from being used for malicious purposes.
Claude Fable 5 is a Mythos-class AI model designed with baked-in safeguards. These restrictions specifically target high-risk domains, such as cybersecurity, to prevent the model from assisting in the creation of digital weapons or exploits. This caution follows a preview of the model earlier this year, which revealed the AI could discover thousands [2] of software vulnerabilities.
While the model is now available globally, the rollout began with limited access for about 200 organizations [1]. These early users include the U.S. government, which is accessing the technology under the Glasswing program [1].
There are conflicting reports regarding the exact nature of the release. Some sources said that Anthropic released a full version of the cybersecurity-centric Claude Mythos model alongside the safer Fable version. However, other reports said that the full Mythos 5 model remains unavailable to the public, and Fable 5 was created as a safeguarded alternative to fill that gap.
Regardless of the specific versioning, the company is prioritizing the prevention of misuse. By restricting sensitive areas, Anthropic aims to provide the utility of a Mythos-class model without the risks associated with unrestricted vulnerability discovery.
“Anthropic released Claude Fable 5 on Tuesday, June 9, 2026”
The launch of Claude Fable 5 represents a growing industry trend toward 'tiered' AI releases, where companies deploy a neutered version of a powerful model to the public while reserving the full capabilities for vetted partners. By citing the model's ability to find thousands of software bugs, Anthropic is acknowledging that frontier AI has reached a level of proficiency where it could potentially destabilize global cybersecurity if left unrestricted.
