Anthropic has confirmed that it accidentally leaked the source code for its popular AI-powered coding tool Claude Code.
The entire source code for Claude Code was accidentally exposed via a misconfigude .map file in its npm package. Over 500,000 lines of code were exposed, revealing all kinds of internal details and features that have never been made public.
Among other things, Anthropic appears to be testing a new mode called “Proactive mode” where Claude Code can be used to code around the clock. There’s also a “Dream” mode that involves Claude thinking and solving problems while you sleep or do something else. And then you have the interactive Tamagotchi-esque “Buddy” feature, which provides a kind of emotional support while you code.
No sensitive customer data or login credentials were leaked. This was a packaging issue caused by human error, not a security breach. In a statement to BleepingComputer, Anthropic says it’s doing what it can to make sure this is a one-time occurrence: “Earlier today, a Claude Code release included some internal source code. No sensitive customer data or credentials were involved or exposed. This was a release packaging issue caused by human error, not a security breach. We’re rolling out measures to prevent this from happening again.”
The leaked source code has already been mirrored numerous times and continues to circulate online. Anthropic’s lawyers are working hard to stop its spread, citing copyright laws. The leak comes at the absolute worst time: just as Anthropic has been enticing users over from other platforms like ChatGPT, just as the company has been surging in popularity and establishing itself as a major leader.
