Anthropic inadvertently triggered the removal of thousands of code repositories on GitHub during an effort to erase copies of its leading product’s source code from the internet.
Earlier this week, a software engineer found that Anthropic had, apparently unintentionally, made the source code accessible for its top-tier Claude Code command line application in a recent update. AI aficionados subsequently analyzed the exposed code, seeking insights into how Anthropic utilizes the foundational LLM, and disseminated it across GitHub.
Operating under U.S. digital copyright law, Anthropic submitted a takedown request to GitHub, seeking the removal of repositories that hosted the infringing code. GitHub’s documentation indicates this notice impacted approximately 8,100 repositories, a count that included authorized forks of Anthropic’s own publicly available Claude Code repository, as noted by aggrieved social media users whose code was consequently blocked.
Boris Cherny, Anthropic’s head of Claude Code, subsequently stated that the action was unintentional and proceeded to withdraw the majority of the takedown requests, confining the scope to a single repository and 96 forks containing the inadvertently exposed source code.
An Anthropic spokesperson informed TechCrunch, stating, “The repository specified in the notice belonged to a fork network linked to our official public Claude Code repository, which caused the takedown to impact more repositories than was our intention.” The spokesperson added, “We have since revoked the notice for all but the designated repository, and GitHub has reinstated access to the impacted forks.”
This mishandled remediation effort represents an additional setback for the company, especially as it reportedly prepares for an IPO—a process that typically necessitates meticulous attention to operational execution and regulatory adherence. The prospect of a public company leaking its source code almost certainly foreshadows an impending shareholder lawsuit.