Anthropic Set to Launch Claude Mythos AI Model to the Public

Claude Mythos Model Goes Public

According to НВ — Техно: Anthropic is preparing to release its Claude Mythos model to the public. An update to the company’s Project Glasswing application reveals plans to broaden access to this AI system. However, the full public rollout of Mythos-class models will only occur once adequate safeguards against misuse are established. Claude Mythos was first introduced in April of this year.

Testing Reveals Security Vulnerabilities

During evaluations, the new AI model demonstrated the ability to generate working exploits in 72.4% of cases, raising significant security concerns. Claude Mythos analyzed over one thousand open-source projects and identified 23,000 bugs, 6,000 of which were deemed critical. A particularly notable discovery was a vulnerability in the wolfSSL cryptographic library, which is used on billions of devices. This flaw allowed attackers to forge certificates and impersonate legitimate websites, though it has since been patched.

In an official blog post, Anthropic representatives stated that Mythos-level models will become widely available within the next 6 to 12 months. However, they also openly acknowledged uncertainty about the timeline, emphasizing that no organization currently has fully developed safety mechanisms in place. For context, the Claude Opus 4.6 model’s exploit generation rate was near zero, highlighting the leap in capability-and risk-with the new system. This information was published on May 28, 2026.

The public release of Anthropic’s Claude Mythos model carries major implications for AI safety, as the vulnerabilities uncovered could threaten numerous users and systems relying on cryptographic solutions. This underscores the urgent need for robust protective measures before deploying new technologies. The anticipated availability of the model within 6 to 12 months signals the company’s active efforts to enhance both security and AI performance.