Anthropic's Claude Code Source Code Leaked

A routine software update for Anthropic’s Claude Code coding tool inadvertently exposed its source code this week. The leak occurred after the release of Claude Code’s 2.1.88 update on Tuesday, revealing features currently under development.

Codebase Briefly Publicly Available

According to reports, a file within the update contained the app’s source code. Before Anthropic addressed the issue, the codebase was uploaded to a public GitHub repository and copied over 50,000 times. The exposed code comprised more than 512,000 lines and 2,000 TypeScript files.

'Proactive' Mode and Potential Payment System Discovered

Alex Finn, founder of AI startup Creator Buddy, discovered a flag for a feature called “Proactive” mode within the leaked code. This mode would allow Claude Code to operate and assist users even without explicit prompts.

Autonomous Payments in Development?

Finn also reported finding evidence of a potential crypto-based payment system. This system could enable AI agents to make autonomous payments, suggesting advanced capabilities for the coding tool.

Virtual Companion Feature Hinted At

In a separate discovery, another individual found indications that Anthropic may have been experimenting with a Tamagotchi-like virtual companion. This companion would “react to your coding,” potentially as an April Fools’ joke.

Anthropic Responds to the Leak

“A Claude Code release included some internal source code. No sensitive customer data or credentials were involved or exposed,” an Anthropic spokesperson stated. “This was a release packaging issue caused by human error, not a security breach. We’re rolling out measures to prevent this from happening again.”

Future Plans Subject to Change

It’s important to note that features identified in leaked code are not guaranteed to be released. As with any software development process, plans are subject to change and features may be altered or abandoned before a final product is launched.