Anthropic debuts a preview of Mythos’ powerful new AI model in a new cybersecurity initiative

Anthropic on Tuesday released a preview of its new frontier model, Mythos, which it says will be used by a small group of partner organizations for cybersecurity businesses. In a previously leaked note, the AI ​​startup described this model as the “most powerful” to date.

Anthropic said the model’s limited debut is part of a new security initiative, dubbed Project Glasswing, in which more than 40 partner organizations will deploy the model for the purposes of “defensive security work” and to secure critical software. Although not specifically trained in cybersecurity work, the inspector will be used to scan first-party and open source software systems for code vulnerabilities, the company said.

Anthropic claims that over the past few weeks, Mythos has identified “thousands of zero-day vulnerabilities, many of which are critical.” The company added that many of the vulnerabilities are one to two decades old.

Mythos is a general-purpose model for Anthropic’s Claude AI systems that the company claims has strong programming and logical thinking skills. Anthropic’s frontier models are their most advanced, high-performance models, designed for more complex tasks, including agent building and coding.

Partner organizations previewing Mythos include Amazon, Apple, Broadcom, Cisco, CrowdStrike, Linux Foundation, Microsoft, and Palo Alto Networks. As part of the initiative, these partners will eventually share what they have learned from using the model so that the rest of the tech industry can benefit from it. Anthropic said previews will not be publicly available.

Anthropic also claims to have been involved in “ongoing discussions” with federal officials about the use of Mythos, though one has to imagine that these discussions are complicated by the fact that Anthropic and the Trump administration are currently locked in a legal battle after the Pentagon described the AI ​​lab as a supply chain risk due to Anthropic’s refusal to allow independent targeting or surveillance of American citizens.

News of Mythos was originally leaked in a data security incident reported by Fortune magazine last month. A draft blog about the model (then called “Capybara”) was left in an unsecured cache of documents available in a publicly inspectable data lake. The leak, which Anthropic later attributed to “human error,” was originally spotted by security researchers. “Capybara is a new name for a new level of model: larger and more intelligent than our Opus models — which were, by far, our most powerful,” the leaked document said, later adding that it was “by far the most powerful AI model we have ever developed,” according to the report.

In the leak, Anthropic claimed that its new model went far beyond the performance areas (such as “software coding, academic thinking, and cybersecurity”) catered by its current generic models, and that it would likely pose a cybersecurity threat if weaponized by bad actors to find and exploit bugs (rather than fix them, which is how Mythos would be deployed).

Last month, the company accidentally exposed nearly 2,000 source code files and more than half a million lines of code through a mistake it made in the release of version 2.1.88 of its Claude Code software package. The company then accidentally took down thousands of code repositories on Github while trying to clean up the mess.