Quick Facts
- Anthropic accidentally leaked details about Claude Mythos, its most capable AI model to date, through an unsecured data store containing nearly 3,000 assets
- The model found over 500 high-severity vulnerabilities in open-source projects and dramatically outperforms Claude Opus 4.6 on coding and cybersecurity tests
- Cybersecurity stocks dropped 3-7% following the leak, with CrowdStrike falling 7% and Palo Alto Networks declining 6%
Anthropic’s most advanced AI model was accidentally revealed through a data leak on March 27, exposing details about Claude Mythos and its unprecedented cybersecurity capabilities.
Security researchers Roy Paz from LayerX Security and Alexandre Pauwels from the University of Cambridge discovered the leak in an unsecured, publicly searchable data store. The cache contained nearly 3,000 previously unpublished assets, including draft blog posts about the new model.
Claude Mythos represents what Anthropic calls “a step change” in AI performance. The model dramatically outperforms Claude Opus 4.6 on software coding, academic reasoning, and cybersecurity tests. Anthropic engineers used the model to find more than 500 high-severity exploits in open-source projects that had gone undetected for decades.
“We’re developing a general purpose model with meaningful advances in reasoning, coding, and cybersecurity,” an Anthropic spokesperson said. “Given the strength of its capabilities, we’re being deliberate about how we release it.”
The leaked documents reveal concerning cybersecurity implications. The draft blog post describes Mythos as “currently far ahead of any other AI model in cyber capabilities” and warns it “presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders.”
Anthropic plans to release Mythos only to select early access customers for cybersecurity defense use cases. The company will not make it publicly available in the short term due to its extreme capabilities and high operational costs.
Markets reacted swiftly to the news. CrowdStrike dropped 7%, Palo Alto Networks declined 6%, and Zscaler fell 4.5%. Other cybersecurity stocks including Okta, SentinelOne, and Fortinet lost 3%. The iShares Expanded Tech-Software Sector ETF dropped nearly 3%.
The leak comes as Anthropic considers going public as early as October 2026 at a $60 billion valuation, according to Bloomberg. The timing also coincides with OpenAI finishing pretraining its newest model, internally called Spud, which is expected to launch within weeks.
Anthropic has experienced explosive revenue growth, scaling from zero to $100 million in 2023, then to $1 billion in 2024. CEO Dario Amodei recently stated the company’s revenue has reached above $4 billion in the first half of 2026.
The accidental disclosure represents one of the most significant unintentional AI model reveals to date, highlighting the growing tension between advanced AI development and cybersecurity implications.
Read more: Anthropic to launch new ‘Claude Mythos’ model with advanced reasoning features