The Pentagon Hopes to Speed Up ‘Kill Chain’ AI Buys With 5-Minute Videos

WIRED reports that a Defense Department procurement program called Tradewinds, run by the Chief Digital and Artificial Intelligence Office, accepts product pitch videos no longer than five minutes. A panel reviews submissions at least monthly. Accepted products are deemed post-competitive, which lets government buyers check the competition box and, per a DOD official, has allowed some awards in under a week. OpenAI, Anthropic and Google are listed as participants.

This year's focus areas include AI to enhance force lethality and AI agents that assist joint targeting, including kill chain execution. Requests from the last Biden-era cycle for tools promoting responsible AI and guardrails are no longer on the list. Acceptance also makes it easier to use other transaction agreements, contracts exempt from some federal acquisition rules; by law the Pentagon can spend up to $500 million on a single one without notifying Congress.

We understand why the Pentagon wants this. WIRED lays out the history: post-Cold War consolidation left the military with few suppliers, and today private AI investment dwarfs what the government can spend, which costs it leverage over where the technology goes. Months-long procurement cycles are a real problem when software changes every few weeks. Bringing new companies in faster is a reasonable goal, and some of what flows through Tradewinds, such as military health tools, is plainly useful.

But look at what is being sped up. The focus areas explicitly include lethality and AI agents in the targeting process. The previous cycle's request for responsible-AI guardrails has dropped away. The contracting vehicle it opens the door to can move large sums with limited reporting; WIRED notes OT spending is hard to find on federal spending sites, and Senator Joni Ernst cites reviews flagging questionable costs, weak oversight and reduced transparency. Put those pieces together and the fastest lane in defense procurement is being pointed at the most consequential use of AI, with the least visibility.

That is the trajectory that worries us. Faster buying is not a harm by itself. Faster buying of targeting tools, judged first on a short video, with guardrail requirements de-emphasized and spending hard for the public to trace, raises the odds that systems reach the field before anyone outside the building can ask whether they work as claimed or fail safely.

The AI labs are part of this story too. OpenAI, Anthropic and Google all declined to comment to WIRED, and all three hold OT agreements with CDAO, though they did not confirm whether Tradewinds was used for them. Companies that publish safety principles should be willing to say how those principles apply when the customer is asking for kill chain support.

Our read: modernizing procurement is fair; doing it by trimming transparency and guardrails on lethal AI is not. I'm watching whether the new public reporting rules on OTs actually surface this spending, and whether any lab draws public lines on targeting work.

Tradewinds has existed since late 2022, but the current administration has been far more vocal about buying and operationalizing military AI, and the White House's 2027 budget request asks for $1.5 trillion for the Defense Department with AI listed as a priority. The Stop Secret Spending Act of 2025, backed by Senator Ernst, will require agencies to report OTs publicly, with three years to comply.