What happened with Anthropic’s Claude Code leak?

Anthropic issues takedown requests after Claude Code source leak

Anthropic has been dealing with an unexpected exposure of Claude Code source code after a leak was traced to packaging and distribution mistakes. Coverage says Anthropic issued copyright takedown requests to remove more than 8,000 copies of the exposed code.

The incident is tied to the way Claude Code was distributed, including cases where a misconfigured package or map file effectively exposed internal TypeScript source information. Reports also describe the leak as occurring at an “inconvenient” moment for Anthropic, given that the product and its capabilities are a major part of its market momentum.

What Anthropic reportedly did next

• Sent copyright takedown requests aimed at removing widely mirrored copies.
• Characterized the leak as a human error / release packaging issue rather than a security breach in some accounts.

Why it’s a big deal

Even if the leaked material doesn’t include the underlying AI model weights, exposing agent tooling can still be strategically important. Source code can reveal internal architectures, operational logic, and integration approaches—information that can help rivals, researchers, or attackers understand product behavior.

There’s also a practical downstream risk: once code is public or widely replicated (including via mirrors), takedown requests may reduce further spread but can’t reliably reverse all prior copies. That makes rapid response crucial.

For enterprises evaluating agentic tools, incidents like this raise the bar for supply-chain hygiene, build/release controls, and how vendors prevent accidental exposure via dependencies or distribution artifacts.

Overall, the leak has moved from an internal release mistake into a broader cleanup operation involving large-scale takedowns.