Restricted Model Moves Toward Public Access
Anthropic is signaling a significant shift in its approach to Claude Mythos, the company’s most advanced AI model that was previously held back from public release. The model first emerged in late March 2026 after an accidental leak from Anthropic’s content management system exposed internal assets describing the model as a major leap in capability. Rather than launching it broadly, the company initially restricted access through Project Glasswing, a defensive cybersecurity initiative that gave exclusive use to more than 40 partner organizations including major technology companies and financial institutions.
Within weeks of deployment through Project Glasswing, Mythos identified over 10,000 high and critical severity vulnerabilities across widely used open source software. This rapid success has apparently prompted a reassessment of the model’s availability, with internal documents now suggesting a staged release to broader enterprise customers.
Integration with Claude Code and Security Dashboard
Evidence of the commercial rollout is emerging through technical signals. Source code strings recently discovered reference a new model identifier, claude-mythos-1-preview, along with language indicating access would be available through Claude Code and a revamped Claude Security dashboard. A screenshot shared by Testingcatalogue briefly showed the Mythos 1 model label visible in the Claude user interface before it was removed, confirming active preparation for release rather than theoretical planning.
An updated Project Glasswing report further supports this direction, stating that Mythos class models could reach the public once appropriate safeguards are in place. This marks a notable departure from earlier positioning that suggested the model would remain indefinitely restricted, pointing toward Anthropic’s confidence in the security controls now being developed for broader deployment.
Source: Cyber Security News
