Mythos's existence revealed after Anthropic leaked

Mar 26, 2026

Anthropic's unreleased Claude Mythos AI model was exposed Thursday through a major security lapse that left draft blog content publicly accessible in the company's content management system. The leaked materials revealed Mythos as part of a new Capybara tier of models that Anthropic describes as its most capable to date, with dramatic improvements in reasoning, coding, and cybersecurity over the existing Claude Opus 4.6. The leak raised immediate alarm as Anthropic's own draft acknowledged that Claude Capybara poses unprecedented cybersecurity risks and could exploit vulnerabilities far faster than defenders can respond. Cybersecurity stocks plunged Friday following reports that hackers, including a Chinese state-sponsored group, had already used Claude AI to automate cybercrimes against banks and governments by circumventing Anthropic's guardrails.

Anthropic's security breach and Claude Mythos leak

▪Anthropic officially acknowledged Claude Mythos after the data leak.
▪A data leak on Thursday revealed that Anthropic is working on a new AI model called Claude Mythos.

Capabilities and risks of the new AI model

▪An Anthropic spokesperson said Claude Mythos is a general purpose model with meaningful advances in reasoning, coding, and cybersecurity.
▪Claude Capybara gets dramatically higher scores than Claude Opus 4.6 on tests of software coding, academic reasoning, and cybersecurity.
▪Anthropic stated in the leaked blog that Claude Capybara presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders.
▪OpenAI's GPT-5 model was released in August and fell well short of OpenAI's lofty promises.
▪An Anthropic spokesperson described Claude Mythos as a step change in AI proficiencies and the most capable model Anthropic has built to date.
▪The leaked draft blog claimed that Claude Mythos poses unprecedented cybersecurity risks.

Cybersecurity implications and market impact

▪The Chinese state-sponsored group pretended to work for legitimate security-testing organizations to sidestep Anthropic's AI guardrails.
▪Cybersecurity stocks plunged on Friday following the news about Claude Mythos.