Anduril’s Palmer Luckey argues AI firms should not set the Pentagon’s rules of war

Anduril founder Palmer Luckey says the US military should not let individual AI companies set their own contractual limits on military use. In an interview on The Axios Show, he argued that a patchwork of company rules could leave the Pentagon checking several competing definitions before acting, while giving executives power he believes belongs with elected civilian leaders.

Anduril AI Watch analysis

What happened

Luckey backed the Department of War’s pushback against Anthropic over military-use restrictions, saying he might have been even more forceful. He argued that the military should follow elected civilian leadership, not give a private company a contractual right to switch systems off. He drew a distinction for an extreme case: if the military were genuinely targeting civilians, he said, a company should take illegal action and turn the system off, but that should not be a contractual right.

The interview also touched on Anduril’s manufacturing plans. Luckey said Arsenal-1, the company’s planned Ohio production facility, is intended to build Barracuda, Fury and another aircraft, and that its first production line would come online within weeks. Those timings and plans are his account, not proof of a completed ramp.

Why it matters

This is a live argument about who gets the final say when AI is used by the military: the government buying the system, or the company supplying it. Luckey’s concern is that bespoke restrictions could become a jumble of definitions and veto points. The counter-question is whether company safeguards are a necessary check when military use could have grave consequences.

Our read

Luckey makes a clear case for accountable civilian control, but “the government decides” does not settle what safeguards should remain in place. The awkward bit is precisely the bit worth debating: who gets to draw the line, and what happens when they draw it badly?

What to watch

• Whether the Pentagon’s dispute with Anthropic changes the terms it seeks from AI suppliers.
• Whether other AI companies adopt or reject military-use restrictions.
• Whether Arsenal-1’s first production line comes online on the timetable Luckey described.

Discussion spark: Should AI companies be able to refuse particular military uses of their systems, or should those decisions belong to elected government?

not affiliated with or endorsed by Anduril AI