The Pentagon is pushing AI into its weapons — and testing a line that used to be red
The Pentagon is accelerating AI across military systems, including weapons able to pick their own targets. In Congress, some senators are trying to slow it down.
For years, the unwritten rule of American defence was simple: the machine assists, the human decides. That line is getting fainter. The US administration is pushing the armed forces to speed up the adoption of artificial intelligence — including, according to the American press, in weapons systems with the autonomy to pick and attack targets on their own, the step that was always treated as taboo.
How much is the Pentagon spending on AI weapons?
The budget request for 2026 includes 13.4 billion dollars just for AI-enabled autonomous systems, 9.4 billion of it for unmanned aircraft. The top brass is no longer hiding the direction of travel: the chairman of the Joint Chiefs said this spring that autonomous weapons will be a key part of the American way of war. The Pentagon has meanwhile set up a unit to consolidate autonomous-warfare technology, reporting directly to the department’s number two.
What would the Responsible AI Defense Act do?
It is the brake part of Congress is trying to pull: a group of senators introduced a bill that would force the Pentagon to classify military AI systems by risk level and subject the most dangerous to extra review before they can be developed or fielded. Critics argue the department’s policy is not keeping pace with the technology — and that life-and-death decisions should not belong to an algorithm. The bill’s text is on the official Congress portal.
None of this is entirely new in Washington — the current leadership’s taste for pushing limits was already on show when the Pentagon decided to screen its troops’ testosterone. But this time the stakes are not blood tests: it is who pulls the trigger. And once a machine answers that question, there is no undo key.
By Oliver Grant
Image: David / Wikimedia Commons (CC BY 2.0)
