Transcription
So, you've probably heard that Anthropic's AI coding tool, Claude Code, just leaked its own source code. But you probably haven't heard about the guy that used AI to rewrite the whole thing just a few hours after the code was leaked.
Here's a story. A developer took the entire leaked codebase, around 512,000 lines of code, fed the code into AI, and got back a full Python rewrite, same features, different language, and still live on GitHub.
Anthropic's legal team came swinging with DMCA takedowns. The original repos went down fast, but the Python version is still up.
And here's the problem no one is talking about. If Anthropic argues the rewrites violate their copyright, they're implying that AI-generated code is protected IP. The same argument could be used against them for training on everyone else's code. And if Anthropic's own engineers don't write their code, which they've claimed, then the leaked codebase may not be protected by copyright at all.
They built a tool that leaked itself and someone with AI rewrote it in an afternoon. And now every legal argument that Anthropic makes cuts both ways. The safety-first AI company just walked into a trap they can't lawyer their way out of.