Anthropic seems to be preparing Mythos 1 for Claude Code.

TestingCatalog reports that Anthropic may be working on a public or broader preview of Mythos 1 for Claude Code and Claude Security. The indications come from leaked UI signals, source code strings, and recent updates regarding Project Glasswing.

Mythos is Anthropic’s very powerful model class for coding, agent tasks, and cybersecurity. In Project Glasswing, Anthropic uses Claude Mythos Preview with selected partners to find vulnerabilities in critical software and help resolve them. According to Anthropic, more than ten thousand serious or critical vulnerabilities have been found in the process so far.

The new signals point to a model name claude-mythos-1-preview and text referring to access to Claude Mythos in Claude Code and Claude Security. Claude Security would also reportedly receive a dashboard for discovered vulnerabilities, historical charts, and triage results.

It is important to note: Anthropic has not yet officially launched Mythos 1 for Claude Code or Claude Security as a product. TestingCatalog bases this on leaked and temporary visibility in interfaces. However, it aligns with Anthropic’s own statement that it intends to make Mythos-class models more widely available in the long term, once stronger safeguards are in place.

This news is relevant because Mythos illustrates exactly the tension surrounding powerful AI coding models. On the one hand, such models can help defenders make software more secure faster. On the other hand, similar capabilities can help attackers find and exploit vulnerabilities more quickly.

For developers and security teams, this means that AI security tools will likely become more professional soon. Not only code assistants, but complete dashboards for vulnerability discovery, scan history, and triage could become part of the development workflow.