Code Leak of Claude Code: Over 512,000 Lines Made Public

An unexpected information leak has unfolded as a result of the update to version 2.1.88 of the Claude Code tool, created by the AI-focused company Anthropic. Approximately 512,000 lines of code from this programming tool, which is designed for artificial intelligence-based programming, have become accessible to a wide audience. This incident has generated significant interest among programmers and technology experts, as the leak has revealed new features, AI instructions, and details about the system's operation, while user data remains untouched.

According to information provided by The Verge, users noticed that the update package contained a file with source code written in TypeScript. An active user of the platform X discovered this issue and decided to share the code file publicly. Reports from Ars Technica and VentureBeat indicate that the leak includes a vast amount of code, allowing for an in-depth understanding of the tool's functionality.

Those who have begun analyzing the code have uncovered numerous new features, internal instructions for the bot, and information regarding its memory system. Notably, the code mentions a virtual 'pet' reminiscent of a Tamagotchi that reacts to user actions while coding. Additionally, a feature named KAIROS is introduced, aimed at ensuring the agent operates continuously in the background, which could significantly enhance the tool's usability.

Furthermore, the code reveals a comment from one of the developers stating that the use of memoization complicates the system, and he is unsure whether it truly improves performance. This indicates that even during the development process, doubts can arise regarding the effectiveness of certain solutions.

Anthropic quickly responded to the situation, rectifying the issue that led to the leak. However, despite their efforts, users managed to copy the code onto GitHub, where it amassed over 50,000 copies of the repository. This suggests that interest in the tool and its capabilities remains high, especially after Claude Code was launched in February 2025 and gained popularity due to new features that allow it to perform tasks on behalf of users.

In a statement, the company noted that the release inadvertently contained part of the internal code, but no confidential data or user accounts were disclosed. This is an important point, as user data security remains a priority for the company operating in the field of artificial intelligence.

Thus, this code leak serves as a crucial lesson for Anthropic and other developers working on similar technologies. It underscores the necessity of adhering to high security standards and vigilance during software development, particularly in the fast-evolving landscape of artificial intelligence technologies.