The Claude Mythos model identified 271 vulnerabilities in Firefox.

BlockBeats News, April 22 — According to Decrypt, Mozilla, the developer of the Firefox browser, published a blog post on Tuesday stating that an early version of Anthropic’s Claude Mythos AI model helped identify 271 vulnerabilities in Firefox during internal testing, which have been fixed this week.

The AI model has recently attracted attention for its performance in the cybersecurity field. Mozilla wrote: “As these capabilities reach more defenders, many other teams are now experiencing the same dizzying feeling we initially saw in the test results. For a hardened target, just one such vulnerability would be enough to trigger a red alert in 2025, and the appearance of so many vulnerabilities at once makes one wonder if it’s even possible to keep up.” Mozilla previously tested another model from Anthropic, which identified 22 security-sensitive vulnerabilities in a previous version of Firefox. Mozilla also stated that the vulnerabilities found did not exceed the capabilities of “elite human researchers.”

Mythos was launched in March this year and is Anthropic’s most advanced model for reasoning, coding, and cybersecurity tasks, belonging to a new tier beyond its earlier Opus series. Tests before the model’s release showed it could identify thousands of previously unknown vulnerabilities across major operating systems and web browsers. Anthropic restricts access to this system through a limited project called Project Glasswing, opening it only to specific tech companies such as Amazon, Apple, and Microsoft. On Monday, reports indicated that despite the Trump administration’s request for federal agencies to cease using Anthropic technology, the NSA is still running Claude Mythos Preview on classified networks.