On March 26, 2026, a misconfigured content management system left a folder of roughly 3,000 Anthropic internal assets publicly accessible online. Among them: a draft launch blog post for a model the company had not yet announced. The model is called Claude Mythos. Anthropic has since confirmed its existence — and the details that have emerged are some of the most significant in the company’s history.
What the Leak Revealed
The accidentally exposed documents included a near-complete draft blog post describing Claude Mythos in detail. Key findings from the leaked materials:
- Mythos represents a new product tier above Opus — internally referred to as “Capybara” — making it the fourth tier in Anthropic’s lineup, above Haiku, Sonnet, and Opus
- Anthropic describes the model as “by far the most powerful AI model we’ve ever developed” and a “step change” in performance
- It delivers dramatically higher scores than Claude Opus 4.6 on software coding, academic reasoning, and cybersecurity benchmarks
- The model is currently being trialed with a small group of early access enterprise customers
- Mythos is described as extremely compute-intensive and expensive to run — Anthropic is working to optimize it before any general release
The Cybersecurity Warning
The most striking part of the leaked materials isn’t the capability claims — it’s what Anthropic says the model can do in the wrong hands. The draft blog post states that Mythos “presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders.”
Anthropic has separately been privately briefing top government officials that Claude Mythos makes large-scale cyberattacks significantly more likely in 2026, according to Axios reporting. The concern is specific: the model is particularly capable at identifying and exploiting software vulnerabilities autonomously. Anthropic’s existing Claude Opus 4.6 has already been used internally to find more than 500 high-severity exploits in open-source projects — including one case where the model deduced the presence of a flaw in a PDF tool by analyzing a developer comment in its change log. Mythos is reportedly substantially more capable at this class of task.
The disclosure triggered a sharp sell-off in cybersecurity stocks — the iShares Expanded Tech-Software Sector ETF dropped nearly 3% following the reports.
Anthropic’s Response
The company confirmed the model in a statement to Fortune: “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. As is standard practice across the industry, we’re working with a small group of early access customers to test the model. We consider this model a step change and the most capable we’ve built to date.”
The deliberate rollout approach — early access to specific organizations rather than a public launch — reflects the company’s stated intent to give enterprises time to harden their defenses against the class of capabilities the model introduces. The draft blog post framed it directly: Anthropic is releasing Mythos in early access specifically to give organizations “a head start in improving the robustness of their codebases against the impending wave of AI-driven exploits.”
What the New Tier Means for Anthropic’s Product Line
If Mythos ships as described, it changes Anthropic’s competitive position significantly. The current Claude Opus 4.6 is already considered one of the strongest models for coding and reasoning. A model that performs dramatically better on those benchmarks would land at a moment when Anthropic is competing directly with OpenAI’s GPT-5.4 series, Google’s Gemini 3.1 Pro, and an accelerating field of open-weight competitors.
The pricing implications are also notable. Mythos is expected to be more expensive than Opus — adding a premium tier above Anthropic’s current ceiling. For developers and enterprises choosing between frontier models, the addition of a new capability tier reshapes the comparison across the entire competitive landscape.
Conclusion
Claude Mythos is the most significant unreleased AI model story of 2026 — revealed not by a planned launch but by a misconfigured server. Whether the general release is months away or longer depends on Anthropic’s optimization work and its assessment of the safety implications. Browse our directory to follow Claude and every major AI model development as they unfold.
