Anthropic Accidentally Leaks 512,000 Lines of Claude Code Source, Exposing Unreleased Features and Future Roadmap

Anthropic accidentally exposes 512,000 lines of Claude's internal source code via an npm package map file, revealing unreleased features, secret codenames, and a more powerful Claude 4.6 variant — marking the company's second major accidental disclosure in two weeks and raising serious concerns about the safety-focused AI firm's own operational security.

• Anthropic confirms that roughly 512,000 lines of Claude Code's internal source code have been accidentally leaked via a map file in an npm package, with human error — not a security breach — cited as the cause, and no customer data exposed.
• The leaked files reveal unreleased features and codenames including Kairos, Buddies, and Capybara, a more powerful Claude 4.6 variant, giving competitors and bad actors a detailed look at Anthropic's proprietary technology and future product roadmap.
• Anthropic moves quickly to issue copyright takedown requests targeting over 8,000 GitHub copies of the leaked code, but the incident marks the company's second major accidental disclosure in two weeks, raising serious reputational concerns for a company built on a safety-first reputation.