Overview
- Anthropic’s Claude Code command-line tool source was accidentally exposed, and AI community members quickly analyzed it and reposted copies on GitHub.
- To curb the spread, the company filed U.S. copyright takedown requests that mistakenly swept up about 8,100 repositories, including legitimate forks and some of Anthropic’s own open-source branches.
- Anthropic said the overreach occurred because the targeted repository sat inside a public branch network that caused the removal scope to balloon across connected branches.
- After user backlash, the company withdrew most notices and GitHub restored access for the vast majority of affected projects, with one repository and 96 related branches still subject to action.
- Claude Code lead Boris Cherni called the exposure a human error, and the incident now carries reputational and legal risks as Anthropic prepares to go public.