Anthropic has built its public identity around the winning idea of being the thoughtful AI company. It publishes detailed papers on the risk of artificial intelligence, employs some of the best researchers in the field, and has voiced the responsibilities of building such powerful technology — so heavy, of course, that it’s currently battling it out with the Department of Defense. On Tuesday, unfortunately, someone there forgot to check a box.
It is, in fact, the second time in a week. Last Thursday, Fortune was mentioned that Anthropic had accidentally released nearly 3,000 internal files, including a draft blog post describing a powerful new model the company had yet to announce.
Here’s what happened on Tuesday: When Anthropic released version 2.1.88 of its Claude Code software package, it accidentally included a file that exposed nearly 2,000 source code files and more than 512,000 lines of code — essentially the complete architectural blueprint for one of its most important products. A security researcher named Chaofan Shou noticed almost immediately and was published on X. Anthropic’s statement to multiple stores was nonchalant, saying: “This was a packaging issue caused by human error, not a security breach.” (Internally, we’re guessing things were less measured.)
Claude Code is not a by-product. It’s a command-line tool that allows developers to use Anthropic’s artificial intelligence to write and edit code, and it’s become formidable enough to annoy adversaries. According to the WSJ, OpenAI pulled the plug to its Sora video production product just six months after its public release to refocus its efforts on developers and businesses — partly in response to Claude Code’s growing momentum.
What was leaked was not the AI model itself but the software scaffolding around it — the instructions that tell the model how to behave, what tools to use, and where its limits are. Developers began posting detailed reviews almost immediately, with one describing the product as “a production level developer experiencenot just a wrapper around an API.”
Whether this proves to be important in any lasting way is a question that should be left to the developers. Competitors may find architecture instructive. at the same time, the field is moving fast.
Either way, somewhere at Anthropic, you can imagine that a very talented engineer has spent the rest of the day quietly wondering if they still have a job. One can only hope it’s not the same engineer, or team of engineers, from late last week.
Techcrunch event
San Francisco, California
|
13-15 October 2026
