Anthropic’s Claude Mythos has drawn fresh attention after some users claimed they briefly saw the model inside Claude Code and Claude Security. The reported appearance came one day after Anthropic said it was unlikely to make Mythos available to the general public soon.
The timing raised questions among developers and security watchers. Anthropic has promoted Claude Mythos Preview as a powerful cybersecurity model under Project Glasswing, but it has also warned that wider access needs stronger safeguards.
Some users on X reported seeing models labeled “Mythos 1” and “claude-mythos-1-preview” inside Claude Code. TestingCatalog also reported that new strings appeared in the source code, including “Access to the Claude Mythos model in Claude Code and Claude Security.”
However, the reported labels appeared only for a brief period. Anthropic has not announced a public launch of Mythos 1, and there is no official confirmation that the model is ready for wider access. For now, the sighting remains an alleged product leak.
Claude Security is Anthropic’s enterprise cybersecurity tool. The appearance of Mythos-related strings suggests the company may be testing how the model could work inside its developer and security products. Still, testing does not mean a public rollout is imminent.
Anthropic has said Claude Mythos Preview remains under restricted access. The company is working with selected partners, including major technology firms and infrastructure developers, through Project Glasswing.
According to Anthropic, Mythos-class models may become available more widely only after stronger safeguards are ready. The company wrote, “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.”
This signals the company’s cautious position. Anthropic appears to see Mythos as useful for defense, but risky if released without strict controls. The model can support vulnerability research, code review, and security testing, yet those same skills could raise misuse concerns.
Project Glasswing has become the main reason Claude Mythos is receiving attention. In one month, more than 50 major developers and infrastructure partners reportedly worked with Claude Mythos Preview. The model helped identify over 10,000 high-severity or critical software vulnerabilities.
Reports linked the work to major systems, including Cloudflare, Firefox, OpenBSD, and security libraries used across connected devices. Some reports also said Mythos found a long-hidden OpenBSD bug that had remained unnoticed for 27 years.
The model’s role was not limited to finding flaws. Reports said it also helped block a $1.5 million cryptocurrency-related fraud attempt at a partner bank. The case reportedly involved suspicious behavior linked to a wire transfer, which Mythos helped detect before funds moved.
These claims have not all been described in full technical detail. That is expected in security reporting, as teams often withhold details until patches are ready. However, Anthropic’s update points to a new phase where AI systems can find bugs faster than many teams can fix them.
The speed of Mythos has created a new pressure point for developers. Finding bugs once took teams weeks or months. Now, AI-assisted tools can scan large codebases and produce detailed reports much faster.
However, fixing vulnerabilities still requires human review, testing, and safe deployment. Some maintainers have reportedly struggled with the number of findings. One reported reaction from the open-source community was, “Please stop finding vulnerabilities—we can’t keep up.”
Anthropic has responded by developing tools that support both discovery and patch creation. Claude Security aims to help enterprise users identify flaws and generate fixes. Meanwhile, the company has also supported responsible disclosure and controlled access for trusted researchers.
The brief Mythos sighting on Claude Code has fueled speculation, but it does not prove that Anthropic has changed its release plan. The company still presents Mythos Preview as a restricted model aimed at defensive cybersecurity work.
For now, the clearest reading is that Anthropic may be preparing Mythos for controlled use inside Claude Code and Claude Security. However, general public access still appears uncertain, especially as Anthropic continues to stress safeguards, partner limits, and responsible security use.
Also Read: Claude Mythos Delay Sparks Debate on Compute Limits and AI Safety Concerns