Anthropic Leaks Source Code of Claude Code in Second Security Blunder This Week

TLDR

• Anthropic unintentionally exposed the internal source code for its AI programming assistant, Claude Code
• The incident revealed approximately 1,900 files containing 512,000 lines of code
• A social media post on X featuring the code accumulated more than 30 million views
• Anthropic has confirmed that no user data or login information was compromised
• This marks the second data-related issue for Anthropic within a week

(SeaPRwire) –   On Tuesday, Anthropic acknowledged that it had inadvertently made public a portion of the internal source code for Claude Code, its AI-driven coding assistant. The firm characterized the event as a “release packaging problem resulting from human mistake, rather than a security intrusion.”

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

Cybersecurity experts reported the leak involved about 1,900 files and 512,000 lines of code. Because Claude Code operates within developer environments and can access confidential data, the exposure alarmed security professionals.

A post on X that linked to the leaked code rapidly gained traction. By Tuesday morning, it had already surpassed 30 million views.

Developers started analyzing the code to gain insights into Claude Code’s functionality and Anthropic’s future development plans. Meanwhile, several security specialists expressed apprehension about the potential misuse of this information by malicious actors.

In a blog post, the AI cybersecurity company Straiker cautioned that attackers could now examine the data flow within Claude Code’s internal pipeline. They indicated this knowledge could enable the creation of persistent payloads during extended sessions, potentially establishing a backdoor.

A Second Incident in One Week

This was not a standalone occurrence for Anthropic. Mere days before, Fortune had reported that the company had mistakenly left thousands of files open to public access.

These files contained a draft blog post detailing a forthcoming AI model, referred to internally as “Mythos” and “Capybara.” The draft allegedly mentioned that the model carries cybersecurity threats.

Anthropic announced it is implementing new procedures to avoid similar incidents. The company emphasized that neither event involved sensitive customer information or credentials.

Claude Code by the Numbers

Anthropic launched Claude Code for public use last May. The tool assists software developers with creating features, resolving bugs, and automating processes.

The assistant has experienced rapid expansion. By February, its run-rate revenue had exceeded $2.5 billion.

This success has prompted rival companies to respond. OpenAI, Google, and xAI have all allocated resources to develop competing coding tools.

Anthropic was established in 2021 by former OpenAI executives and researchers. The company is primarily recognized for its suite of Claude AI models.

A spokesperson for the firm stated that measures are being taken to prevent a recurrence of such a leak.