Claude Mythos Leak Sparks Fresh AI Security Concerns

Several U.S. cybersecurity stocks fell on Friday, with the market tying the sell-off to rumored leaks involving Anthropic’s new model, Claude Mythos. Public reports said descriptions related to the model were stored in a publicly accessible and searchable data cache ahead of release, resulting in the exposure of draft content. On the day, CrowdStrike fell about 7% and Palo Alto Networks fell about 6%, while other names such as Okta also came under pressure.

The leaked draft claimed Anthropic plans to introduce a new model tier called Capybara, potentially larger and more capable than its current flagship Opus, and described Claude Mythos as “the most powerful yet.” The draft also warned that the model could significantly outpace others in cyberattack and vulnerability exploitation capabilities, potentially triggering a new round of competition in which “defenders may struggle to keep up.” Investors therefore reassessed whether, if AI can automate threat detection and response at scale, some premium-priced security products could face incremental substitution by “AI-native capabilities” and increased pricing pressure.

Commentary: Near-term sentiment could amplify volatility across the sector, but enterprise security remains highly complex in terms of compliance, delivery, and accountability boundaries. The key next steps will be Anthropic’s handling of the leak and further disclosures on the new model’s real-world capabilities and deployment scenarios.