Anthropic's Claude Code source code accidentally leaked on GitHub
Anthropic accidentally leaked the source code for its Claude Code AI agent on GitHub on March 31, 2026. The code spread rapidly and generated millions of views before Anthropic issued copyright takedown notices to remove it. Security experts noted the exposed framework could help competitors understand Anthropic's engineering approach, though critical core elements were not compromised.
What happened
Anthropic accidentally exposed the source code for its AI agent Claude Code on GitHub on March 31, 2026. The code spread rapidly, accumulated millions of views, and spawned multiple derivative versions on GitHub before Anthropic took action.
Anthropic issued Digital Millennium Copyright Act (DMCA) takedown notices based on copyright to remove the leaked code and its forks from GitHub repositories.
Reported impact
A cybersecurity expert noted that while no critical core elements were exposed, the leaked code revealed the 'harness'—the software framework connecting Claude Code's language model to practical applications—which competitors could use to understand Anthropic's approach to difficult engineering challenges.
Index analysis
The incident exposed a paradox in the AI industry: the same tools enabling rapid product development also make information leakage and replication instantaneous and effortless, regardless of confidentiality status.
The leak created an ironic situation where Anthropic relied on copyright law to protect its own work while facing multiple lawsuits from creators alleging the company trained its models on their copyrighted content without permission.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.