Anthropic has deleted thousands of GitHub repositories in an attempt to exfiltrate its leaked source code — a move the company says was an accident

Anthropic accidentally took down thousands of code repositories on GitHub while trying to pull copies of the source code for its most popular product from the Internet.

Anthropic has, seemingly by accident, included access to the source code of its class-leading Claude Code command-line application in a recent release, a software engineer discovered Tuesday. AI enthusiasts mined the leaked code for clues on how Anthropic could harness the LLM material underlying the app, and shared it on GitHub.

Anthropic issued a takedown notice under U.S. digital copyright law asking GitHub to remove repositories containing the infringing code. According to Github logs, the notice was implemented against about 8,100 repositories — including legitimate forks of Anthropic’s Claude Code repository that were released publicly, according to angry social media users whose code was blocked.

Boris Cherny, head of Cloud Code at Anthropic, said the move was accidental and withdrew the bulk of the takedown notices, limiting them to a single repository and 96 forks with source code released in error.

“The repo mentioned in the notice was part of a subnet connected to our public Claude Code repository, so the takedown reached more repositories than intended,” an Anthropic spokesperson told TechCrunch. “We pulled the notification for everything except the one repo we named, and GitHub has restored access to the affected forks.”

The botched cleanup here is another black eye for the company as it reportedly plans an initial public offering, a task that typically requires attention to enforcement and compliance. Leaking your source code as a public company? You better believe there’s a shareholder lawsuit coming.