Anthropic inadvertently released the code behind Claude AI this week, which the internet quickly noticed and exploited. While Anthropic has not issued a public statement on the FUBAR event, it did release a statement to multiple news organization acknowledging the error:
“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.”
According to WSJ.com, Anthropic saw over 8,000 copies of the code shared on GitHub. The company sought to have these removed by issuing legal requests, but it is doubtful that this will be effective.
One post on X claimed that a Korean developer, who is said to be an active Claude dev, rewrote the code in Python, then Rust, in an attempt to evade the copyright requests. Chatter indicates the code will be available forever.
The error now exposes details of Claude AI that may be helpful to competitors. There may also be a security risk, at least until the code is rewritten and updates are made available.
Claude is estimated to have over 18.9 million monthly active users, and its estimated monthly web traffic reached 176.12 million by December 2025. The platform has grown in popularity for creating autonomous agents capable of operating a computer.