New Delhi: Anthropic is racing to contain a leak involving its AI coding tool, Claude Code, after internal source code briefly became public and spread
across developer platforms. The company has sent copyright takedown notices to attempt to take the material down, but the code has already spread far and wide, so it is difficult to contain.
The leak has generated a wave of interest amongst programmers, with many of them busy preserving and republishing the code. Although, according to Anthropic, no sensitive user data was disclosed, the event has created an interest in the safety of intellectual property in the rapidly advancing AI sector.
Claude code source code has been leaked via a map file in their npm registry!
Code: https://t.co/jBiMoOzt8G pic.twitter.com/rYo5hbvEj8
— Chaofan Shou (@Fried_rice) March 31, 2026
Leak origin and company response
A report by PCMag states that the problem started when Anthropic accidentally shipped internal source code in a 59.8MB file as part of a now-deleted release of Claude Code (version 2.1.88). The company subsequently admitted the error, claiming it was a release packaging accident due to a human error and not a security breach.
Anthropic was first advocating broad-scale removals, and GitHub has been reported to have been handling takedown requests for thousands of repositories. The company, however, reduced its request later on in order to concentrate on a few links. These measures did not help since copies of the leaked code were already distributed on the platform.
Developers evade takedowns
The developer community has been resisting the efforts to eliminate the code. It is also reported that programmers are modifying the leaked material by rewriting it into other scripting languages like Python and Bash. Such a strategy complicates the task of copyright protection systems to identify and delete the content.
According to some developers, when such code is publicly available, it is almost impossible to completely wipe out. A consultancy mentioned in the news reports said the material would be permanently public even when original sources are removed.
What the leak reveals
Initial indications indicate that the leak does not contain sensitive assets such as model weights, training data or backend infrastructure. Nevertheless, it does provide some understanding of the way Claude Code works. Developers who looked at the files have claimed to have found capabilities meant to deter illegal cloning and systems that mask indications of AI-generated content.
The accident may also affect the competitive advantage of Anthropic. The leak can provide competitors in the AI world with important information on how to enhance their own code generators by revealing details of the internal architecture of the company.
