Staff Reports
The Pentagon’s senior tech official threw down the gauntlet this week: we do not need Congress to rush out a sweeping new law to regulate artificial intelligence. Under Secretary of War for Research and Engineering and Department of War Chief Technology Officer Emil Michael told a national TV audience that product liability and existing executive tools already give the government the teeth it needs. He was blunt, and he was right to warn against panic-driven lawmaking that could do more harm than good.
Pentagon: Treat AI like a product, not a political theater
On a recent broadcast, Under Secretary of War for Research and Engineering and Department of War Chief Technology Officer Emil Michael said the right approach is simple: treat AI systems like other products and hold companies accountable when their products cause harm. He pointed to civil and criminal penalties, and to familiar legal doctrines like product liability. Michael even used the Ford Pinto example to remind listeners that the courts can punish reckless manufacturers — no taxpayer bailout required.
Anthropic and export controls: the practical test
Michael cited the Anthropic episode earlier this year as proof that the executive branch can act fast when a model looks dangerous. When regulators restricted access to certain advanced models, Anthropic disabled the models while talks followed. That sequence shows how export‑control and national‑security tools can be used to constrain a risky release without Congress writing an emergency statute. The White House’s voluntary frontier‑model review framework also gives agencies a pre‑release window to assess new systems, which is part of the toolbox Michael was describing.
Counterpoints — and why they don’t erase the core argument
Critics say the law hasn’t kept up and want Congress to set bright-line rules. CEOs and many lawmakers argue for statutory guardrails, and insurers worry about how to price systemic AI risk. Those are real issues. But the alternative — slapping a one-size-fits-all statute on a fast-moving field — risks freezing innovation or handing vague power to regulators who love regulation. Michael’s point is worth remembering: don’t let fear push us into hasty laws that could “turn us into Europe,” as he put it. Better to use liability, market pressure, and targeted executive action where they work.
Where to go from here
Supporters of strong AI governance should press for clarity, not chaos. Congress can do useful, narrow fixes: clarify liability standards, strengthen insurance markets, and codify limited authorities like export controls so companies and courts know the rules of the road. But the reflexive demand for sweeping new regulation is the wrong answer. Let the market, the courts, and the executive branch use the tools they have — and save broad, heavy-handed legislation for problems that actually need it.
