tech
April 2, 2026
Anthropic has mistakenly leaked its own internal Claude code
Anthropic has reportedly exposed internal code for its Claude AI system due to a release error, raising concerns over potential misuse

TL;DR
- Anthropic accidentally published its top-secret internal design and code for its Claude Code AI assistant.
- The incident was attributed to human error, not a security breach.
- Over 500,000 lines of code, including unreleased features and developer notes, were leaked.
- The leaked code spread rapidly online and was replicated on platforms like GitHub.
- The exposure could potentially allow malicious actors to bypass security or trick the AI system.
- Anthropic previously faced scrutiny from the US Defense Secretary over its AI systems.
- A separate leak in February exposed details of Anthropic's unreleased, powerful Claude Mythos model.
Continue reading the original article