Anthropic Accidentally Leaks Claude’s Source Code

[Narrator] A major security issue

at Anthropic has people talking,

after the full source code

for Claude Code was accidentally leaked to the public.

And no, this is not an April Fool's joke.

By Tuesday morning, the leak was still accessible

and security researcher Chaofun Shou reported it

and flagged it publicly on X.

He discovered that the file pointed

to a zip archive on Anthropic's own cloud storage

containing the full source code with nearly 2000 files

and 500,000 lines of code.

Based on the code shared on GitHub,

the leak appeared after an apparent internal debugging file

was accidentally published

as part of a regular Claude Code update.

From there, it spread quickly.

The code was rapidly backed up on GitHub,

where a repository has already been

forked over 41,500 times,

spreading it widely and turning Anthropic's slipup

into a resource for the AI and cybersecurity world.

While parts of Claude Code have previously

been reverse engineered,

this is the most widespread view into the AI system so far,

and it's not the first Anthropic leak either.

Earlier this year,

a misconfigured data store

exposed nearly 3000 internal files,

including unpublished blog posts, PDFs

and memos that reveal details

of Anthropic's unreleased Claude Mythos model.

Anthropic confirmed the leak in a statement

to Business Insider saying the release had unintentionally

included internal source code,

but no sensitive customer data or credentials were exposed.

The company described it as a packaging error caused

by human mistake, not a security breach,

and said it's implementing measures to prevent a repeat.