Why federal agencies are secretly testing Anthropic's new 'Mythos' model: Here's what it means

Anthropic frontier AI model 'Mythos' is undergoing significant changes in line with its cyber proficiency

Federal agencies are reportedly bypassing a Trump administration directive to test Anthropic's most advanced AI model, Claude Mythos. The primary context behind this revelation is that the Commerce Department's Center for AI Standards and Innovation is vigorously examining Anthropic's frontier AI model, Mythos, hacking prowess. Claude Mythos is a model currently in development mode and being tested by Anthropic, positioned above the existing Opus series.

Anthropic has confirmed that this model represents a leap in capability and is the most powerful system the company has built to date. According to leaked drafts, the model scores significantly higher than Claude Opus 4.6 in critical areas, including software programming and academic reasoning.

Over the past week, three congressional committees held or requested certain briefings from the company to learn about Mythos's cyber scanning capabilities over the past week. In response, Anthropic's co-founder Jack Clark said at the Semafor World Economy Summit on Monday that the company is continuing to discuss Mythos with the Trump administration following the Pentagon's decision to cut ties with the US AI firm in the wake of an ongoing dispute.

The model's performance in cybersecurity is described as far surpassing all existing AI models, marking a major technical milestone for the industry. This information surfaced after unpublished drafts and assets were exposed due to a configuration error in Anthropic's content management system. Because of its advanced capabilities and high operational costs, Anthropic reportedly plans to offer the model initially to a select group of early customers specifically on cybersecurity defense. The leak involved nearly 3,000 unpublished assets, including images, PDFs, and internal documents. At present, the nature and details of Anthropic's talks with the US government remain unclear, including which specific agencies are involved.

Mythos, which was officially announced on April 7, is Anthropic's "most capable yet for coding and agentic tasks" highlighting its ability to act autonomously. Due to the model's extreme power and high-operational cost, Anthropic will not release it to the general public in the short term. Instead, it will be offered only to a select group of early customers specifically for cybersecurity defense use cases.