From the Department of You-Can't-Make-This-Up:

After a heated back-and-forth between Defense Secretary Pete Hegseth and the Defense Department's leading artificial intelligence supplier Anthropic over the firm's insistence that its systems not be used for mass surveillance or autonomous warfare, President Donald Trump banned the government's use of Anthropic's Claude AI tool.

A few hours later, the United States attacked Iran — reportedly with the help of Claude.

Among the first casualties was an elementary school where more than 165 people, mostly kids, were reported killed, a school adjacent to an Islamic Revolutionary Guard Corps naval base. Claude is being used in part for "target identification," leading to questions about whether it was used in this strike and if so, how its judgment was off. Meanwhile, fake AI-edited videos and other images regarding that incident and others went viral, a new iteration of the fog of war.

After blacklisting Anthropic, Hegseth signed a deal with OpenAI, whose AI systems the government had considered inferior to Anthropic's. Criticism from OpenAI personnel and public blowback forced the company to belatedly amend the agreement to include protections against mass surveillance.

A roundup of highlights from Newsday's Opinion Department.

By clicking Sign up, you agree to our privacy policy.

This followed closely the release of a study by King's College in London which found that when advanced AI models from Anthropic, OpenAI and Google were placed in war games simulations, they threatened or opted to use nuclear weapons in 95% of the scenarios.

If all of this makes you queasy, it should. AI continues to be the elephant in the room, even if no one understands exactly what kind of elephant it is going to be.

Poll after poll shows Americans don't trust AI. Further raising the stakes: AI money is now pouring into politics. In last week's primaries, a super PAC affiliated with OpenAI invested at least $500,000 each in three new Republican House candidates and each won their race. All three are in ruby-red districts, pretty much guaranteeing the election of three new pro-OpenAI lawmakers.

If you're in the market for a silver AI lining, there might be one in the Hegseth-Anthropic battle: It was the first major debate held in public about where to place limits on AI. But even that scintilla of progress is tarnished. It has long been generally accepted that AI must be regulated. In September 2023, major tech executives including Elon Musk, Mark Zuckerberg, Bill Gates and Sundar Pichai, along with leading AI advocates and skeptics at a summit convened by then-Senate Majority Leader Chuck Schumer, unanimously agreed that the federal government should regulate AI. Two and a half years later, crickets. Congress, as usual, can't get it done.

That's no surprise to the public. People don't believe government can capably regulate AI and don't trust tech companies to self-police. Both positions are understandable.

Surveys consistently capture that discontent. Large majorities agree that AI will reduce jobs, that it poses a threat to humanity, and that companies can't be trusted to use AI responsibly. Some 90% of women worry about the lack of AI regulations to protect children, and huge swaths of Americans are concerned about surveillance, fake images, AI starting wars — and more profoundly, the possible rule of tech and a loss of control of everyday life.

For those who despise elites and already feel powerless, AI looms as one of the biggest baddies of all.

Politicians are beginning to respond to these fears. Witness the sudden retreat by those who had championed the big data centers essential to AI development. But data center opposition is only a proxy for the more fundamental conflict. 

What's happening with AI is quintessentially modern American: The public is deeply worried about Big Tech and unconvinced that their leaders can deal with the issue, as those leaders prove by their inability or refusal to act that those fears are justified.

The stasis is alarming, as AI races on.

Columnist Michael Dobie is a retired member of the Newsday editorial board.

Michael Dobie is a retired member of the Newsday editorial board.

© Newsday