Anthropic is testing its latest AI model, "Claude Mythos," which follows a data leak revealing potential cybersecurity risks. The leak exposed nearly 3,000 unpublished assets due to a CMS misconfiguration. A spokesperson declared Mythos the company's "most capable" model to date, featuring a new tier, "Capybara," designed for enhanced reasoning and coding. Anthropic warns it may enable advanced cyberattacks, prompting initial release plans aimed at cyber defenders.