Claude Code Leak: 8100 Takedown Requests and the Birth of Claw-Code (www.heise.de)

Anthropic faced a significant scandal when a human error led to the accidental public exposure of over half a million lines of source code for its Claude Code, a programming assistance tool. This leak, which occurred when a crucial source map file was mistakenly published, resulted in the swift dissemination of the code online, prompting Anthropic to issue over 8,100 takedown requests under the DMCA to halt its spread. The leak is notable not just for revealing code but for disclosing the architecture behind autonomous agents in AI, which are capable of making independent decisions and adapting their actions based on user interactions, offering competitors valuable insights for system improvements. In response to the takedown frenzy, the developer community quickly launched Claw-Code, an independent Python rewrite of the leaked architecture, attributed to developer Sigrid Jin with assistance from OpenAI's Codex. Claw-Code has already become the fastest-growing repository in GitHub's history, signaling a shift in how AI-generated code and copyright are viewed. This situation raises complex legal questions about the ownership of AI-generated content, especially since much of Claude Code was produced by the AI itself. As the traditional concepts of copyright struggle to keep pace with AI advancements, the incident may signal the decline of secretive intellectual property protections in favor of more open-source models within the software industry.