admin
The Department of War’s top tech official just told viewers what some in Silicon Valley seem unwilling to admit: the idea of a quick, grand bargain with China to “slow down” frontier AI is unrealistic and risky. Under Secretary of War for Research and Engineering Emil Michael said on Fox News that the notion of a treaty to handcuff American advantage is “extremely naive” and “crazy.” That blunt pushback should shape any talk of an AI treaty, and it matters for national security, tech policy, and upcoming diplomatic meetings.
Pentagon Chief Technologist Rejects an AI Treaty with China
Emil Michael, the Department of War’s chief technology officer, answered a question on live TV about industry calls to “pace the frontier” and coordinate with China. He said that if the United States is ahead, accepting a deal where the U.S. slows while China benefits is nonsense. His warning hits at the heart of Dario Amodei’s essay urging industry and governments to adopt pacing and international checks on AI. China’s own foreign ministry quickly pushed back on Amodei’s plan, calling it “fearmongering,” so the diplomatic ground for a swift treaty is thin.
Why a Treaty Would Be Self-Handicapping and Hard to Verify
Pacing sounds pretty when you’re reading it in a think-piece. In reality, it means one side trusting another not to sprint while it jogs. Verification is a headache. How do you inspect sensitive models, chips, and research in a country that sees strategic advantage in domination? The Pentagon’s view is simple: you don’t ask to be tied up if your edge wins wars, protects troops, and keeps deterrence strong. That’s not paranoia. It’s basic strategy.
Industry Naivete Meets Real-World Security
This clash is also personal. The Pentagon and frontier labs like Anthropic have already been in tense talks over contracts and terms. When a defense buyer needs reliability and security, vague promises about “global coordination” don’t cut it. Congress is watching, export controls are on the table, and procurement rules can force labs to choose: support defense needs or lose business. Saying “let’s trust everyone” won’t survive a hearing in front of lawmakers worried about chips, models, and battlefield use.
Conservative readers should like Michael’s straight talk. We shouldn’t torch American advantage in the name of idealistic accords that China will ignore or exploit. That does not mean wild-west AI with no rules. It means focused guardrails: export controls, clear procurement terms, and tough verification inside U.S. alliances — not a naive, one-size-fits-all treaty. If the tech industry wants safety, fine. But safety must be realistic, enforceable, and aligned with national security. Otherwise, “pacing the frontier” will look less like wisdom and more like handing the competition a head start.
