Claude Code Source Leaked via npm Packaging Error, Anthropic Confirms
Why it matters: Anthropic is issuing copyright takedown requests for over 8,000 copies of its leaked Claude Code source.
- Anthropic confirmed the accidental leak of Claude Code's internal source code via an npm packaging error, assuring no sensitive customer data was exposed.
- The Verge AI and TechMeme (The Information) revealed the leaked code includes a Tamagotchi-style 'pet,' an always-on agent, and 'Kairos' updates like a 'dream mode' for consolidating memories.
- BBC Tech reported that Claude Code users are hitting usage limits 'way faster than expected' following the leak.
- Sam Schechner/Wall Street Journal noted that Anthropic is issuing copyright takedown requests to remove over 8,000 copies of the leaked code.
- Stratechery contextualized the Claude Code leak alongside other security incidents, including the Axios Supply Chain Attack.
Anthropic confirmed its AI coding assistant, Claude Code, had its internal source code inadvertently leaked due to a human error during npm packaging, though no sensitive customer data was exposed. The leak, which Anthropic is aggressively trying to contain with takedown requests, revealed features like a Tamagotchi-style 'pet,' an always-on agent, and 'dream mode' for memory consolidation, leading some users to experience faster-than-expected usage limits.
