TL;DR
- Claude Code: Anthropic appears to have exposed Mythos in public Claude Code surfaces before a spotted toggle vanished again.
- Release Path: Mythos still sits behind restricted Project Glasswing access while Anthropic weighs a wider rollout.
- Security Stakes: The model promises stronger code reasoning and vulnerability work, but misuse controls remain unresolved.
- Market Context: Amazon and Windsurf already ship broad coding tools, raising the bar for any Mythos release.
Anthropic, appears to have moved Mythos closer to public use after Mythos appeared in public Claude Code references on May 25 and users saw a Mythos toggle before it disappeared again. Mythos has so far remained inside a restricted preview, and is not available via mainstream coding products.
Anthropic first introduced Mythos Preview on April 7 as a frontier model for advanced computer security tasks. Claude Code now shows the clearest public sign yet that Anthropic is weighing how to release Mythos-class models more widely without opening them fully.
ANTHROPIC 🔥: Mythos 1, “claude-mythos-1-preview”, is being prepared for a release on Claude Code and Claude Security.
The model became visible for a short amount of time on Claude; besides that, new strings mentioning Mythos have been added.
> Access to the Claude Mythos… pic.twitter.com/ig16Gx1Zsw
— 🚨 AI News | TestingCatalog (@testingcatalog) May 23, 2026
Why the Claude Code Sighting Matters
Public rollout plans still look unsettled. Anthropic kept Mythos outside general customer access while it built stronger controls against misuse, so a visible Claude Code appearance signals testing at the edge of a broader release rather than a finished launch.
Claude Code users would feel the difference quickly. Even a limited rollout would put a cyber-focused model closer to everyday software work while Anthropic tries to preserve the controls that limited the model’s initial release to critical industry partners and open-source developers through Project Glasswing.
Anthropic is still deciding how to widen access. Nothing in the public traces confirms broad availability yet.
What Anthropic Says Mythos Can Do
Mythos delivers code reasoning gains over Opus 4.7. For developers, that points to stronger help on debugging, security review, and longer chains of coding work that require the model to hold more context and make more autonomous decisions.
Mythos can identify and exploit zero-day vulnerabilities across major operating systems and browsers. Previously unknown flaws are valuable to defenders who want to fix weak points, but they are just as valuable to attackers who want working attack paths.
Risk rises because the same model improvements also increase exploit-generation capability. Better reasoning is useful for legitimate engineering tasks, yet it can also shorten the path from finding a bug to building a functional exploit.
Anthropic’s Glasswing update adds the operational constraint. Verification, disclosure, and patching still take longer, which means defensive teams may remain slower to confirm and fix weaknesses than automated systems are at uncovering them.
Glasswing Scale and Prior Context
Glasswing already shows Mythos operating in controlled defensive work. About 50 partners used Claude Mythos Preview to find more than 10,000 high- or severe-severity vulnerabilities across widely used software.
Scale on that level makes the Claude Code sighting more consequential than a cosmetic interface test. Mythos has already been used to surface serious flaws in restricted environments, so any move toward public developer tools would change how close the model is to everyday product exposure.
Anthropic now faces a harder balancing act. Glasswing proves the model can help defensive teams at scale, while the Claude Code traces suggest pressure to turn that work into a broader product.
