Anthropic Claude Mythos Leak Reveals Next-Gen Model With "Step Change" Capabilities — CMS Misconfiguration Exposes 3,000 Internal Assets Including Cybersecurity Risk Assessment

On March 26-27, 2026, Fortune journalist Bea Nolan discovered a significant data breach at Anthropic — a misconfigured content management system had left nearly 3,000 internal documents publicly accessible on the company website. Among the exposed assets were draft announcements for Claude Mythos, described as Anthropic's most capable model to date.

THE MODEL: Two draft versions were found — one using the name "Mythos" and another using "Capybara" — suggesting Anthropic was deciding between name candidates. Key characteristics from the leaked drafts:

• Represents a new class above Opus: "larger and more intelligent than our Opus models — which were, until now, our most powerful"
• Achieves "dramatically higher scores on tests of software coding, academic reasoning, and cybersecurity, among others" compared to Claude Opus 4.6
• Described as "currently far ahead of any other AI model in cyber capabilities"
• Both name versions use identical justification: chosen to evoke "the deep connective tissue that links together knowledge and ideas"

ANTHROPIC RESPONSE: Anthropic confirmed to Fortune: "We're developing a general purpose model with meaningful advances in reasoning, coding, and cybersecurity. We consider this model a step change and the most capable we've built to date."

The company dismissed the documents as "early drafts of content that were being considered for publication."

CYBERSECURITY CONCERNS: The leaked documents state Mythos "presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders." This aligns with RSAC 2026 warnings from security leaders about AI-accelerated vulnerability exploitation.

ROLLOUT PLAN:

• Deliberately slower release than previous models
• Starting with small early-access customer group for cybersecurity evaluation
• Claude API access expanding gradually
• Acknowledged as "very expensive for us to serve, and will be very expensive for our customers to use"
• Working to make it more efficient before general release

IPO CONTEXT: The leak comes as both Anthropic and OpenAI race toward potential IPOs later in 2026. OpenAI is also preparing its own next-gen model codenamed "Spud" which CEO Sam Altman claims can "really accelerate the economy." The timing suggests both companies are positioning their strongest models for maximum market impact.

MARKET REACTION: The leak caused Bitcoin to slide alongside software stocks. CoinDesk described Mythos as a potential "cybersecurity nightmare" given its stated capabilities in vulnerability discovery and exploitation.

The irony was not lost on observers: Anthropic — the AI safety-focused company that just won an injunction protecting it from Pentagon retaliation — exposed its most sensitive model details through a basic CMS misconfiguration.