Anthropic accidentally deleted thousands of GitHub repositories while attempting to hide code leak

Anthropic found itself at the center of a major scandal: in attempting to remove leaked source code from GitHub, the company mistakenly sent thousands of copyright violation notices — thereby forcibly taking down a vast number of repositories that had nothing to do with the leak. This became one of the largest incidents involving mass DMCA requests in the platform's history. The company acted in standard intellectual property protection mode: it discovered that its proprietary code had appeared in the public domain and attempted to promptly remove it.

However, the automated notification process spiraled out of control — the damage extended not only to repositories containing the leaked materials but to thousands of completely unrelated projects. Anthropic leadership acknowledged the error and stated that everything happened unintentionally. The company rescinded most of the notices it had sent and is working with GitHub to remedy the consequences for affected users.

For developers whose repositories became inaccessible, this meant losing access to their own projects — albeit temporarily. The fact of the code leak itself deserves separate attention. The company develops the Claude family of models — among the most sought-after in the corporate segment — and traditionally keeps its technical developments under the strictest secrecy.

The appearance of code in the public domain indicates a serious failure in the information security system. The response in the form of mass DMCA notices appears to be an attempt to halt the spread of materials as quickly as possible without first assessing the scope of collateral damage. The situation exposes a systemic problem: automated copyright protection tools work quickly, but not always accurately.

Code similarity algorithms can mistakenly identify legitimate projects as violators — especially if the leaked fragment has spread widely across forks, mirrors, and training repositories. When such tools are deployed at scale without sufficient verification, any error in selection criteria becomes a mass failure. GitHub found itself in an ambiguous position.

The platform is obligated to respond to DMCA requests from rights holders in accordance with U.S. law, and it has no technical means to verify each request for accuracy in real time.

The world's largest platform for developers became an unwitting instrument in another company's mistake. This incident raises a broader question about how technology companies should approach intellectual property protection in the era of open platforms. The principle of "block now, investigate later" is tactically understandable, but in practice it leads to collateral damage.

The victims are precisely those developers and researchers who build trust in the company and use its products. Rescinding the notices remedies the immediate consequences but does not address the reputational questions. Anthropic actively positions itself as a responsible player in the AI industry, emphasizing security and trust.

The incident involving mass deletion of others' repositories — even if accidental — contradicts this image and demonstrates that even companies with high standards can experience internal process failures at the most inopportune moments.