Anthropic's DMCA Takedown Accidentally Removes 8,100 GitHub Repos After Code Leak

Anthropic Code Leak Triggers Mass DMCA Takedown

AI startup Anthropic found itself in a difficult position this week after an accidental leak of source code for Claude Code, its AI-powered application for software developers, prompted a DMCA takedown that removed far more than intended. The leaked Claude Code source code first appeared on GitHub when a software engineer discovered on March 31, 2026, that Anthropic had unintentionally included access to the complete source code in a recent release 1

1

. The accidental leak of source code exposed commercially sensitive information, including internal tools, proprietary techniques, and specific instructions for transforming AI models into functional coding agents 3

3

.

GitHub Repositories Takedown Affects Thousands

In response to the leak spreading across developer communities, Anthropic issued copyright takedown requests under US copyright law late Tuesday. The DMCA takedown focused on a repository posted by GitHub user nirholas and nearly 100 specifically named forks 1

1

. However, GitHub processed the notice against an entire network of 8,100 repositories because "the submitter alleged that all or most of the forks were infringing to the same extent as the parent repository" 1

1

. This expanded action swept up numerous legitimate repositories that had forked Anthropic's official public Claude Code repository, which the company shares to encourage public bug reports and fixes 1

1

.

Intellectual Property Protection Effort Backfires

The overzealous intellectual property protection attempt sparked immediate backlash from developers who found their legitimate work blocked. Coders took to social media to voice frustration, with developer Robert McLaws stating, "I'm sorry that your people shipped your source code, and that your lawyers don't know how to read a repo" 1

1

. Boris Cherny, Anthropic's head of Claude Code, acknowledged on social media that the takedowns were "not intentional," while Thariq Shihipar attributed the mishap to "a communication mistake" 1

1

. By Wednesday, Anthropic moved quickly to correct the error, requesting GitHub restrict takedowns to the 96 fork URLs specifically listed and reinstate all other repositories 1

1

.

AI Coding Tools Complicate Containment Efforts

Even with corrected focus, Anthropic faces significant challenges containing the spread of its leaked code. Multiple copies remain easily accessible on GitHub, and the code has appeared on platforms like Germany-based Codeberg, which sits outside direct US DMCA jurisdiction 1

1

. More concerning for the company, enterprising developers have already used AI coding tools to create clean room reimplementations, converting the original TypeScript code into Python and Rust 1

1

. These functionally similar rewrites might be legally distinct, though debate continues over whether AI-written code can claim the strict separation needed to avoid being considered derivative work 1

1

.

AI-Generated Code Copyright Questions Emerge

The incident raises complex questions about AI-generated code copyright. Cherny admitted in December that "in the last thirty days, 100% of my contributions to Claude Code were written by Claude Code" 1

1

. This admission could prove significant, as the US Copyright Office offers protection to "AI-assisted" codebases but generally doesn't extend that protection to work generated entirely by AI 1

1

. The company confirmed the error occurred during release packaging due to human error, not a hack, and stated no confidential client data was exposed 2

2

. As Anthropic reportedly prepares for an IPO while rival OpenAI recently announced a $122 billion fundraise, this incident highlights the growing pains facing AI companies as they balance open development practices with protecting proprietary technology 2

2

.