Anthropic issued 8,000 takedown requests after accidentally leaking Claude AI's full source code The leak exposed proprietary techniques, internal tools, and instructions for Claude's coding agents Leak stemmed from a packaging error, with no customer data or security breach reported by Anthropic