Anthropic’s Restricted Claude Mythos Moves Toward Public Release via Claude Code and Security

In Cybersecurity News - Original News Source is cybersecuritynews.com by Blog Writer

Spread the love

Anthropic appears to be loosening its grip on Claude Mythos, the company’s most powerful and previously restricted AI model, with new signals pointing to a commercially versioned release under the name Mythos 1 (claude-mythos-1-preview), integrated directly into Claude Code and a revamped Claude Security dashboard targeting enterprise customers.

Claude Mythos first surfaced in late March 2026 after an accidental leak from Anthropic’s content management system exposed close to 3,000 unpublished internal assets, including a draft blog post describing the model as “by far the most powerful AI model we’ve ever developed”.

The company confirmed the model’s existence days later, calling it a “step change” in capabilities, particularly in cybersecurity, coding, and academic reasoning.

Rather than a general release, Anthropic launched Project Glasswing on April 7, 2026, a restricted defensive cybersecurity initiative that gave exclusive access to over 40 partner organizations, including AWS, Apple, Google, Microsoft, NVIDIA, CrowdStrike, and JPMorgan Chase.

The model was deployed to scan critical infrastructure, and within weeks had identified more than 10,000+ high- or critical-severity vulnerabilities across widely used open-source software.

Mythos 1: Signals of a Staged Commercial Release

The posture is now shifting. An updated Project Glasswing report notes that the initiative has expanded its protective scope to include open-source projects and states that “Mythos-class models could reach the public once the right safeguards are in place” — a notable departure from earlier framing that held the model as indefinitely restricted.

More concretely, source code strings recently surfaced referencing claude-mythos-1-preview alongside a new access description: “Access to the Claude Mythos model in Claude Code and Claude Security.”

Testingcatalogue shared a screenshot of the Mythos 1 model label in the Claude UI before it disappeared, confirming that the product name is in active preparation rather than in hypothetical planning.

This places Mythos 1 squarely on a path into two products: Claude Code Anthropic’s agentic developer environment and Claude Security, the vulnerability scanning and patch-suggestion product currently in public beta for Enterprise customers.

Claude Security was originally launched in February 2026 as a limited research preview powered by Claude Opus 4.7, scanning codebases and suggesting targeted patches for human review. The current beta is available exclusively to Claude Enterprise customers, with access for Team and Max subscribers announced as forthcoming.

Structural work is now underway on a new dashboard interface that surfaces discovered vulnerabilities alongside seven-day and thirty-day historical charts and deeper triage results.

Evidence in the Claude navigation bar shows a standalone Security tab being built alongside Claude Code and Claude Design, packaging the scanner into a consumer-style dashboard rather than keeping it behind enterprise sales conversations.

The architecture supports periodic automated background scans, on-demand manual scans, and pull requests with proposed patches ready for human review before merging.

Anthropic’s coordinated vulnerability disclosure dashboard, updated as of May 22, 2026, reflects the scale of these efforts: 1,596 vulnerabilities disclosed across 281 open-source projects, with 97 patched to date.

Rumors of Claude Opus 4.8 in internal evaluation with select Anthropic partners have also circulated, which would align with the cadence established by Opus 4.7’s April 16 release.

Anthropic’s own rollout of Opus 4.7 included cybersecurity safeguards that automatically detect and block high-risk offensive use cases framed explicitly as a stepping stone toward a broader Mythos-class release.

The emerging picture is one of deliberate, layered commercialization: Mythos 1 as a gated capability within enterprise products first, followed by measured expansion once safety evaluations and infrastructure safeguards catch up to the model’s formidable and potentially dangerous cyber capabilities.

Follow us on Google News, LinkedIn, and X to Get More Instant Updates.