Anthropic, a prominent AI firm, has confirmed it accidentally exposed the entire source code for its AI-powered coding tool, Claude Code. The significant leak involved over 500,000 lines of proprietary code, revealing numerous internal details and unreleased features of the software.
The exposure stemmed from a misconfigured .map file found within the tool’s npm package. This technical oversight inadvertently made the full scope of Claude Code’s development visible to the public, marking a notable incident in AI software deployment.
Among the revelations were details concerning features that have yet to be publicly released. These insights offer a glimpse into Anthropic’s ongoing development and future capabilities for its coding assistant.
One such feature, dubbed “Proactive mode,” suggests an advanced capability for Claude Code to operate continuously, facilitating coding tasks around the clock without direct constant user intervention.
Another unreleased mode uncovered is “Dream” mode. This feature appears designed to allow Claude to engage in problem-solving and thinking processes independently, potentially while a user is asleep or occupied with other activities, aiming to progress coding projects autonomously.
Furthermore, the leaked code revealed an interactive “Buddy” feature. Described as Tamagotchi-esque, this function aims to provide a unique form of emotional support to developers as they work, integrating a companionable element into the coding experience.
Crucially, Anthropic has confirmed that the incident did not compromise any sensitive customer data or login credentials. The company emphasized that user privacy and security were maintained throughout the accidental exposure.
The firm clarified that the event was a packaging issue resulting from human error, rather than a malicious security breach. This distinction highlights an operational challenge in the intricate process of deploying sophisticated AI software.
In a statement provided to BleepingComputer, Anthropic addressed the situation directly, stating: “Earlier today, a Claude Code release included some internal source code. No sensitive customer data or credentials were involved or exposed. 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.”
Anthropic is actively engaged in efforts to prevent similar occurrences in the future. The company is also working to halt the circulation of the exposed code online, with its legal team citing copyright laws in attempts to manage the spread.
The timing of this leak is particularly sensitive for Anthropic, as the company has been experiencing a surge in popularity, successfully attracting users from competing platforms like ChatGPT and solidifying its position as a major leader in the AI landscape. The incident underscores the complex deployment challenges inherent in the fast-evolving field of AI software development.