So, picture this: you're the CEO of a hot AI startup, and you get a summons to the Pentagon. Not for a casual chat over coffee, but for what one defense official is already calling "not a friendly meeting." That's the situation Anthropic CEO Dario Amodei finds himself in, reportedly set to meet with Secretary of War Pete Hegseth on Tuesday morning to talk about the military's potential use of the company's AI chatbot, Claude.
Here's where it gets interesting. While Anthropic told reporters that their discussions with the Department of War are "productive" and in "good faith," defense officials paint a different picture. They say the negotiations are teetering on the edge of failure, with no real progress made so far. It's the classic corporate "everything is fine" versus the government's "this is about to blow up" narrative.
The core of the tension seems to be about rules. Anthropic is apparently willing to ease up on some of its terms of service, but it's drawing a hard line in the sand on two things: it won't let its tech be used for mass surveillance of Americans, and it won't allow it to be used to develop autonomous weapons. The Department of War, unsurprisingly, views those restrictions as, well, overly restrictive. A report last week even suggested Hegseth was considering cutting ties with Anthropic altogether and slapping the startup with a "supply chain risk" label.
Adding another layer to this already complicated week for Anthropic is a separate, but equally dramatic, saga. The company recently accused three Chinese AI companies of creating a staggering 24,000 fraudulent accounts to exploit Claude. They're calling it the largest documented case of AI model capability theft to date.
Enter Elon Musk (TSLA). Never one to miss a chance to comment, Musk posted on X, "Anthropic is guilty of stealing training data at massive scale and has had to pay multi-billion dollar settlements for their theft. This is just a fact." It's a spicy reminder that in the world of AI, accusations of data theft can fly in all directions.
Oh, and in case you thought Anthropic was only dealing with geopolitical and ethical firestorms, the company also just launched Claude Code Security on Friday. It's an AI tool that autonomously scans codebases for vulnerabilities, with the company claiming its Opus 4.6 model found over 500 previously unknown high-severity flaws in live open-source projects. So, it's been a busy few days.
The Department of War and Anthropic did not immediately respond to requests for comment from MarketDash.
