Meet Claude Mythos: Leaked Anthropic post reveals the powerful upcoming model
Anthropic confirmed details of its powerful upcoming AI model, Claude Mythos, following a data leak that exposed internal assets.Summary
AI company Anthropic confirmed details about its most powerful upcoming model, Claude Mythos, after a significant data leak exposed nearly 3,000 internal assets, including an unpublished blog post draft. The leak, caused by an error in the content management system where files were not marked as private, also revealed information about another unreleased tier called Capybara, which is expected to sit above the current top-tier model, Opus. The draft described Claude Mythos as a "step change" in AI performance, currently in trial with early access customers. Furthermore, the leaked material highlighted Anthropic's concerns regarding the cybersecurity risks posed by Claude Mythos, noting its advanced capabilities in exploiting vulnerabilities and the need to share findings with defenders to prepare for a wave of AI-driven exploits.
(Source：Mashable)