So, this nearly happened, as told in a recent headline: “AI hallucination of Chinese nuclear components almost led to US military attack.” The fact that an artificial intelligence made a possibly grave error is less interesting than the fact that the American military is apparently deploying bad AI in mission-critical circumstances.
The incident plays nicely into the new political mood that seems to have transitioned from “Oh, cool, AI can write my business proposal” to “Oh, no, we’re all extras in Terminator 7.” And that doesn’t even get to what might be the most toxic part of the AI debate for the typical American: new data centers being built in their communities and driving up utility costs. (Oh, yeah, also the bit about AI destroying huge swathes of jobs.)
The world’s politicians and militaries don’t need AI to stumble into wars. Back in the early 1980s, the Soviets nearly launched a nuclear attack on the West, having misinterpreted a NATO military exercise (dubbed “Able Archer) as a preemptive attack. Obviously, war was averted, and most of us never even heard about it until decades later. But there were two salutary effects of the Able Archer near-miss: the U.S. and Soviet Union worked to reduce tensions, and the incident is the focus of a very entertaining German-made TV miniseries (called Deutschland 83).
But the military example does show just how fast and widespread AI’s adoption has become, even for mission-critical tasks. What should be the AI industry’s moment to bask in public adoration has instead been tainted by a series of disturbing news reports, which were underscored when some AI employees recently warned that AI could “kill us all.”
Here are some more headlines to help you sleep at night: “Hackers breached OpenAI, adding to fever pitch of security and safety concerns” (NBC); “Google Says Its A.I. Hacked Three Companies in Testing Breakout” (The New York Times); “AI chatbots give wrong answers to financial queries ‘most of the time’” (Financial Times); “OpenAI reveals concerning new AI behavior and vows to track it more closely” (PBS News). And on and on. You get the point. There’s a lot of stuff going on in the AI world that will change pretty much everything but might be going off the rails and maybe kill us all. But don’t be afraid. (One more headline from BBC: “OpenAI CEO Sam Altman says ‘right to be afraid of AI’ — but ‘world should trust’ tech firms.”)
Trump said he would create an “AI Force” to vaguely deal with the threat — mind you, the threat is a midterm blowout in November.
The American people could use some calming down on AI. In late July, a Gallup Poll found that awareness of AI is quite high, but the percentage of Americans who say AI does more harm than good rose from 31 percent in 2025 to 39 this year. And, pace Sam Altman, Gallup says over the past year that trust in business to use AI has declined over the past year. (Not dramatically so, but still a downward trend.) Other polls are finding high anxiety over the cutting-edge technology (such as a U.C. Berkeley and Politico poll that found 77 percent of voters “said advanced AI is likely a threat to humans”). And on that volatile data center issue, a UMass Poll released last week found that only one in 9 Americans support construction of AI data centers in their communities.
Reading the political moment, governors and legislators across the country (and around the world) have been rolling out data center moratoriums. In the interests of inserting some cooling rods into the overheated reactor of public opinion, or at least to boost his presidential prospects, California Gov. Gavin Newsom has suggested a special legislative session focused on AI. Another of Newsom’s ideas is for a possible “kill switch” to turn off the T1000 if it goes berserk.
One powerful person who opposes tighter regulation of AI is President Donald Trump. Having called concerns over AI a “HOAX,” he later said he would create an “AI Force” to somehow vaguely deal with the threat — mind you, the threat to him is a midterm blowout in November. All we can be sure of is that the AI Force will have a snazzy uniform, because Trump had already declared “The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!”
Will the legislation and time-outs and other attempts to come to grips with AI work? Will the four San Francisco supervisors who have introduced a possible 45-day pause on approving new data centers have any impact, just considering the very likely possibility that this controversy will last more than a month and a half? (A San Francisco waterfront property is already under consideration to host a new AI data center, say reports.)
Because — and mark this on your iCalendar, because it might be the first and only time I write this — Trump is right. In that kinda/mostly/sorta way that is the only way he could be correct about anything. The train has left the station on AI (or, if the worriers are correct, the Hindenburg has left the hanger), and our international competitors are not going to help us out. The Chinese are not going to mothball their successful AI efforts; Russia isn’t going to do anything that would hamper its war against Ukraine. And Europe is moving feverishly to develop its own homegrown AI.
The legislative efforts will likely do little to provide a healthy environment for the growth and sound regulation of AI. Way too many cooks in way too many kitchens. I suspect the biggest impact will be the grassroots efforts to fight new data centers, for which there is a much clearer solution: stop letting big tech companies pay communities in beads and baubles for their water and electricity; Silicon Valley expects to be well paid for its products, services, and resources, so local communities should play that market game, too.
Beyond that, the legislative and snazzy uniforms responses to the AI panic aren’t designed to do much more than to feed into the next election. We’ll get through this the way we’ve learned to get through everything in recent decades: too late, too little, and more of the same.
And don’t expect journalists to be of any help.
Last week CNN’s T.M. Brown decided to ponder, “Will AI really kill everyone? How, exactly?”
Good lord, don’t give it ideas.
