Anthropic’s mysterious cybersecurity-focused AI model, Claude Mythos, may have briefly surfaced publicly just a day after the company suggested it had no immediate plans for a wider release. The accidental appearance sparked fresh speculation online about whether Anthropic is quietly preparing broader access to one of its most closely watched AI systems. The situation unfolded after several users on X claimed they spotted models called “Mythos 1” and “claude-mythos-1-preview” inside Claude Code and Claude Security for a short period before they disappeared again. That timing raised eyebrows because Anthropic had only recently stated that Mythos would remain restricted for now.
Mythos Briefly Appears Inside Claude Tools
According to posts shared by TestingCatalog on X, users noticed references to the Mythos model appearing directly inside Anthropic’s developer tools interface. Some also reportedly found backend source-code strings mentioning: “Access to the Claude Mythos model in Claude Code and Claude Security.”
Claude Security, for context, is Anthropic’s cybersecurity-focused platform, while Claude Code is aimed at software development workflows. The models were visible only briefly before traces seemingly vanished again, which has led many to believe the appearance may have been accidental or tied to internal testing.
Anthropic Recently Denied Public Release Plans
The surprise appearance became even more interesting because Anthropic had just announced on Friday that Mythos would likely remain unavailable to the public for the foreseeable future. The company previously revealed that Claude Mythos had discovered more than 10,000 serious software vulnerabilities in under a month across critical software systems. That alone positioned the model as one of the most powerful cybersecurity-focused AI systems currently known publicly.
Anthropic’s Mythos Found 10,000 Bugs In Weeks. Is The AI Hacking Era Near?
At the same time, Anthropic stressed that stronger safeguards were still needed before wider deployment. “In the near future, once we’ve developed the far stronger safeguards we need, we look forward to making Mythos-class models available through a general release,” the company said.
For now, Anthropic claims Mythos access remains limited to around 40 selected organisations, including Google and Amazon, through something called Project Mythos.
Previous Mythos Sightings Have Also Surfaced
Interestingly, this is not the first time traces of Mythos have appeared unexpectedly online. Reports previously claimed that unauthorised users briefly accessed parts of the model. Other references reportedly surfaced through AWS and Google Cloud vulnerability discovery systems as well. Anthropic later removed those traces too.
