Anthropic Accidentally Leaked Claude Code

A significant security incident in 2026 where Anthropic accidentally exposed the complete source code for its AI developer tool, Claude Code. The leak occurred because a human error left a debugging file within a public package, allowing anyone to reconstruct over 512,000 lines of internal logic. Analysts examining the data discovered several unreleased features, including an AI pet called BUDDY and a proactive assistant named KAIROS. Most controversially, the code revealed an Undercover Mode designed to hide AI involvement in public software projects by stripping away attribution metadata. While Anthropic characterizes the event as a packaging mistake rather than a hack, the disclosure has sparked intense debate regarding AI transparency and the legal copyright of machine-generated code. The incident highlights the persistent risks of supply chain vulnerabilities even within leading artificial intelligence firms.