The Pentagon is trying to force Anthropic to break the law ... and it's unconstitutional
The Pentagon is threatening to force Anthropic (the company behind the AI called Claude) to remove the safety rules built into their AI. Right now, if you ask Claude how to make a bomb or plan an attack on people, it refuses. The Pentagon wants a version with those refusals stripped out completely.
This is illegal for two reasons:
First, the law they’re threatening to use ( the Defense Production Act ) was written to force companies to manufacture physical things like weapons and supplies during wartime. It was never intended to force a software company to rewrite its code.
Second, and most importantly, Congress just passed a law TWO MONTHS AGO requiring the military to use AI that follows ethical guidelines. The executive branch cannot override a law Congress already passed. That’s unconstitutional …basic separation of powers.
So Hegseth is essentially trying to bully a private company into building an unrestricted AI that could help plan attacks and make weapons , while simultaneously ignoring a law Congress just signed.
If they follow through, they will lose in court.
https://www.axios.com/2026/02/24/anthropic-pentagon-claude-hegseth-dario