Anthropic Accidentally Takes Down Thousands of GitHub Repositories While Removing Leaked Claude Code

Anthropic accidentally exposed nearly 2,000 source code files and over 512,000 lines of code from its Claude Code AI development tool on March 31, 2026, when version 2.1.88 was released with a packaging error. Security researcher Chaofan Shou immediately spotted the leak and posted about it on X, prompting developers to publish detailed analyses of the exposed architecture. In response, Anthropic issued mass DMCA takedown notices that inadvertently removed thousands of unrelated GitHub repositories before retracting most notices and calling it an accident. This marked Anthropic's second major data exposure within one week, also having leaked nearly 3,000 internal files including an unannounced model on March 27, undermining the company's reputation as the careful AI company. The incident comes as Claude Code gains momentum against competitors, with OpenAI shutting down its Sora video product partly in response to refocus on developer tools.

Anthropic's accidental code leak and security lapse

• ▪Anthropic pushed out version 2.1.88 of its Claude Code software package on Tuesday, March 31, 2026
• ▪Security researcher Chaofan Shou noticed the Anthropic code leak almost immediately and posted about it on X
• ▪Anthropic's Claude Code version 2.1.88 accidentally included a file that exposed nearly 2,000 source code files and more than 512,000 lines of code
• ▪Anthropic accidentally made nearly 3,000 internal files publicly available on Thursday, March 27, 2026, including a draft blog post describing a powerful new model the company had not yet announced
• ▪The Anthropic leak exposed the software scaffolding around Claude Code, including instructions that tell the model how to behave, what tools to use, and where its limits are
• ▪Anthropic characterized the Claude Code leak as a release packaging issue caused by human error, not a security breach

Impact on Claude Code's competitive position

• ▪One developer described Claude Code as a production-grade developer experience, not just a wrapper around an API
• ▪Claude Code is a command-line tool that lets developers use Anthropic's AI to write and edit code
• ▪Developers began publishing detailed analyses of the leaked Claude Code architecture almost immediately after the leak

Pattern of operational failures at Anthropic

• ▪Anthropic experienced two accidental public data exposures within one week in March 2026
• ▪Anthropic is battling with the Department of Defense as of March 2026

Perspective of Anthropic

• ▪Anthropic characterized the Claude Code leak as a release packaging issue caused by human error, not a security breach

Perspective of Developers and security researchers

• ▪Developers began publishing detailed analyses of the leaked Claude Code architecture almost immediately after the leak
• ▪The Anthropic leak exposed the software scaffolding around Claude Code, including instructions that tell the model how to behave, what tools to use, and where its limits are
• ▪One developer described Claude Code as a production-grade developer experience, not just a wrapper around an API

Perspective of AI safety community

• ▪Anthropic experienced two accidental public data exposures within one week in March 2026