Anthropic's Most Powerful AI Model Exposed: Claude Mythos Data Leak Reveals a New Era in Artificial Intelligence

Anthropic Confirms the Model is Real

Following Fortune's inquiry, Anthropic did not deny the leak. Instead, the company issued a statement validating the core claims.

An Anthropic spokesperson confirmed: "We're developing a general-purpose model with meaningful advances in reasoning, coding, and cybersecurity. Given the strength of its capabilities, we're being deliberate about how we release it. We consider this model a step change and the most capable we've built to date."

The company added that it is currently working with a small group of early access customers to test the model, noting it is expensive to run and not yet ready for general release.

Cybersecurity Risks: A Model That Worries Its Own Creator

Perhaps the most striking aspect of the leak is what Anthropic itself wrote about the risks of the model. Leaked documents suggestClaude Mythos could identify and exploit software vulnerabilities and might enable large-scale cyberattacks if misused.

The draft materials included claims of unprecedented cybersecurity risks associated with the model's capabilities. The language in the draft blog acknowledged that the model "presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders," indicating Anthropic is acutely aware of the dual-use risk embedded in these capabilities.

Scope of the Leak

The data exposure went beyond just the model announcement. In total, close to 3,000 assets linked to Anthropic's blog that had not been published to public-facing sites were nonetheless accessible in the data cache, according to Alexandre Pauwels, a cybersecurity researcher at the University of Cambridge, who Fortune asked to review the material.

The leak also exposed details of a closed-door summit planned for European corporate CEOs. The event was described as an "intimate gathering" at an 18th-century manor in the English countryside, where attendees would hear from lawmakers and experience unreleased Claude capabilities.

After being informed of the exposure, Anthropic removed public access to the data store and confirmed the issue was unrelated to Claude, Cowork, or any Anthropic AI tools.

What This Means for the AI Landscape

The accidental exposure of Claude Mythos gives the industry its clearest signal yet that the next generation of foundation models will not simply be incremental upgrades. With Anthropic openly acknowledging cybersecurity risks serious enough to warrant a cautious, staged rollout, the conversation around AI safety and capability oversight is about to intensify significantly.