t>

Anthropic is one month old


Anthropic has built its social identity on the successful assumption that it is a conservative AI company. It publishes detailed information on the risks of AI, uses the best researchers in the field, and has been talking about the responsibilities that come with building such a powerful technology – clearly, so right now. to deal with it and the Department of Defense. On Tuesday, unfortunately, someone there forgot to check the box.

It is, in fact, the second time in a week. Last Thursday, Fortune report that Anthropic accidentally made nearly 3,000 internal files publicly available, including blog posts detailing new capabilities the company had yet to announce.

Here’s what happened on Tuesday: When Anthropic released version 2.1.88 of its Claude Code software, it accidentally included a file that revealed nearly 2,000 code files and more than 512,000 lines — essentially the construction plans for one of its most important components. A security researcher named Chaofan Shou saw it immediately and wrote on X. Anthropic’s statement to the general store was rather unflattering as far as these things go: “This was a release issue that was caused by human error, not a security breach.” (Inside, we assume that things have not been tested.)

Claude Code is no small thing. It’s a command-line tool that allows developers to use Anthropic’s AI to write and edit code and has been strong enough to disrupt its competitors. According to WSJ, OpenAI he pulled the plug in his video production Sora just six months after launching it to the public to refocus his efforts on developers and businesses – partly in response to the growth of Claude Code.

What came down was not the AI ​​model itself but the software that revolved around it—the instructions that told the model how it should behave, what tools it should use, and what its limits should be. Developers began publishing detailed reviews immediately, with one describing the product as “a. the creativity of the designersnot just covering the API. “

Whether this will be relevant in any way is a question not left to the developers. Competitors can access training infrastructure; at the same time, the field moves quickly.

In any case, somewhere at Anthropic, you can imagine that some very talented engineer spent the day quietly wondering if he still had a job. One can only hope it’s not the same engineer, or engineering team, from last weekend.

Techcrunch event

San Francisco, CA
| |
October 13-15, 2026



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *