Anthropic’s Leaked Drafts Expose Powerful New AI Model “Claude Mythos”

Anthropic inadvertently exposed highly sensitive internal documents—including a draft blog post and risk assessments—about an unreleased AI model named "Claude Mythos" via an unsecured, publicly searchable data cache; the leak highlights both operational security failures (misconfigured access controls) and the potential for significant cybersecurity and reputational impacts because internal assessments reportedly flagged the model as posing elevated cyber risks.