The source code of Anthropic’s developer-specific AI service has been exposed to the public, which the tech community is terming as a massive IP blunder. Chaofan Shou, an intern at Solayer Labs, announced the discovery on X on Tuesday opening floodgates of concerns from members of the tech community.
The source code of Anthropic’s product called “Claude Code” has been exposed in this leak. This product has been launched as a high-end agentic command-line interface (CLI) tool for software engineers.
As per Shou’s update, “Claude code source code has been leaked via a map file in their npm registry!”
This essentially means that Anthropic accidentally failed to remove a technical translation file—known as a source map that ended up exposing the source code of “Claude Code” to the public.
A line-by-line view of over 500,000 Anthropic codes is now all over the internet giving a full peak into how the AI company has shaped up the product.
Shou has shared a link to the exposed code along with a screenshot of how the leak looks.
As of press time, Anthropic has not addressed the situation.
Meanwhile, Shou’s post has rattled the AI industry having garnered over 14.5 million views in the last seven hours.
The screenshots of the revealed code are now making the rounds on social media.
Many are also celebrating that they have managed to mirror this exposed code before Anthropic handles the situation.
Members from the engineering circle are alerting these code copiers against exploiting the situation.
Anthropic’s published npm package contained Claude Code v2.1.88. The source map file cli.js.map that is now leaked contained the full and unobfuscated TypeScript source code.
The discussions around this development on social media are dissecting the coding line-by-line to obtain intricate technical information on Anthropic products.