TLDR Tech / InfoSec
Claude Code Leak: 8100 Takedown Requests and the Birth of Claw-Code
An accidentally published source map file at Anthropic exposed over half a million lines of Claude Code's source code, with security researcher Chaofan Shou first spotting the public directory and tracing the compiled code back to its originals before mirrors spread across GitHub. Anthropic responded with more than 8,100 DMCA takedown requests (later narrowed to roughly 100 specific copies), but Korean developer Sigrid Jin used OpenAI's Codex to produce Claw-Code, a Python rewrite of the agentic framework that became the fastest-growing repo in GitHub history and is reportedly being adopted by xAI. The incident raises a thorny authorship question — Claude Code was reportedly ~90% AI-written, and US courts have held that fully autonomous AI creations don't qualify for copyright protection — leaving Anthropic in the awkward position of enforcing rights on largely machine-generated code while also exposing how brittle the DMCA process is when platforms must remove content without judicial review.
Read the full story