On the ground, the backlash against AI manifests itself as opposition to new data centres, which are accused of hogging electricity. Bills to freeze construction of such centres have been proposed in at least ten states. Yet “the data centre fights are not just about energy, water, land use [and] noise,” says Brendan Steinhauser, a Republican who heads The Alliance for Secure AI Action, a lobby group. “They’re also about AI’s impact on society.”
“The accelerationist or ‘let them cook’ mentality has less influence than it did a year ago,” Mr Steinhauser says. “People want to slow it down and get it right.” What getting AI policy “right” means in practice is unclear. Technology and politics move at utterly different speeds: a new model is born in the time it takes Congress to schedule a hearing. Dozens of AI bills have been drafted, some of them hundreds of pages long, covering everything from tougher safety checks to tighter rules on surveillance. Nothing big has passed yet.
“Most Americans already know AI could be transformative and dangerous at the same time. They’re not confused about that,” says Alex Bores, a computer scientist and New York state assemblyman. “What they want to know is whether anyone in government is actually going to do something about it.” Mr Bores campaigned on the promise that he would, making AI regulation the centre of his bid for Congress in a crowded Democratic primary in New York on June 23rd.
So, some big AI firms tried to crush him. A group funded by tech titans spent millions trying to defeat him for, among other things, wanting a national version of a New York state bill he co-authored that requires AI firms to disclose alarming incidents and not release any model that creates an unreasonable risk of critical harm. Yet another group, funded by other AI firms, spent millions supporting him, precisely because he wants to impose national guardrails. In the end he lost narrowly to Micah Lasher, a more conventional candidate who said that he, too, would stand up to tech firms.
The question of how AI will shape society has become politically explosive in the country where the technology is most advanced. Both Republicans and Democrats have factions that relish the potential rewards of AI, and fear that clumsy regulation will let Chinese models, which are only a few months behind, catch up. Donald Trump’s advisers include David Sacks, who favours developing AI as rapidly as possible. As Mr Bores’s case illustrates, the industry itself is deeply divided. However, voters are not.
Three-quarters of Americans think AI should be more regulated, according to YouGov, a pollster, with Republicans nearly as keen as Democrats. Almost two-thirds of Americans think the technology is advancing too quickly, versus only 2% who say “too slowly”, according to the Pew Research Centre. Policy is struggling to keep up with the pace of change.
Americans of both parties fear that AI will disrupt people’s lives, their jobs and even their children’s future romantic attachments. A whopping 45% say they are very or somewhat concerned that AI might cause human extinction, according to YouGov. Fully 65% think it will reduce the number of jobs available in America (74% of Democrats and 56% of Republicans). More than 70% fret about the exacerbation of mental-health problems and the concentration of power in tech firms.
The intensity of Americans’ anxiety sets them apart. A poll by Pew last year found that Americans were more pessimistic about AI than people in any of 24 other countries surveyed. More recent polling by YouGov finds that 61% of Americans think AI will affect society “a lot”; only 5% think it will affect it only “a little” or “not at all”. For all the promises of riches and new drugs, Americans think that AI’s effects will be more negative than positive, by 46% to 31%. Democrats are slightly—but only slightly—more gloomy than Republicans.
For the progressive left, AI is a hotpot of repulsive ingredients. They worry it will destroy union jobs. They are NIMBYish about data centres. They detest billionaires, and don’t get them started on the trillionaire who pals around with Mr Trump and wants to put data centres in space.
Today’s “tech barons…are far worse than the railroad or coal barons of the ‘Great Gatsby’ age,” says Malcolm Kenyatta, the vice-chair of the Democratic National Committee. “With the development of AI, you have a small group of very rich, mostly white men trying to determine the future for our communities, for our society, for our economy,” says Igor Volsky of Tax the Greedy Billionaires, a pressure group.
Progressives also fret about the effects of AI on human relationships. “I have a ten-year-old. I’m concerned that very soon my child might develop a relationship with an AI chatbot,” says Maurice Mitchell, the leader of the Working Families Party, a left-wing party that endorses Democrats such as Zohran Mamdani. He notes that, in the absence of regulation, some AI tools are marketed “as though they can replace actual therapists [and] actual friends”.
Left-wing laments are echoed by some on the right. Josh Hawley, a Republican senator, is among the most concerned. “The voter backlash is increasing,” he observes. He worries that AI could disrupt 40-50% of entry-level white-collar jobs, and then come for blue-collar jobs, too. It would be fine to use AI to make, say, lorry-driving safer, he says, but he would ban self-driving lorries. In this, he makes common ground with organised labour. On June 16th, at a meeting with the Teamsters union, he accused “mega corporations” of having “lost their moral compass”, reports Axios, a news website. If AI generates fortunes for the few while tossing the rest on the scrapheap, he told The Economist, “I don’t think you can sustain a democracy.”
Like his colleagues on the left, Mr Hawley is also anxious about social effects. Young people are hooked on screens and shunning marriage and parenthood, worries Mr Hawley. AI will “accelerate those trends”. He adds another fear, grounded in faith: that the plutocrats who control AI will try to turn themselves into superhumans, augmented by technology. They may even try to “live infinitely in the cloud”. In his revulsion at this Mr Hawley, a Protestant, sounds a lot like the pope, who recently fretted that “transhumanism” may allow a technological elite to think that some lives are “less worthy” than others.
On the ground, the backlash against AI manifests itself as opposition to new data centres, which are accused of hogging electricity. Bills to freeze construction of such centres have been proposed in at least ten states. Yet “the data centre fights are not just about energy, water, land use [and] noise,” says Brendan Steinhauser, a Republican who heads The Alliance for Secure AI Action, a lobby group. “They’re also about AI’s impact on society.”
“The accelerationist or ‘let them cook’ mentality has less influence than it did a year ago,” Mr Steinhauser says. “People want to slow it down and get it right.” What getting AI policy “right” means in practice is unclear. Technology and politics move at utterly different speeds: a new model is born in the time it takes Congress to schedule a hearing. Dozens of AI bills have been drafted, some of them hundreds of pages long, covering everything from tougher safety checks to tighter rules on surveillance. Nothing big has passed yet.
The president can act more quickly, but has been inconsistent. Last year the Trump administration sought to quash state regulation on AI by executive orders and legislative tricks. A White House AI legislative framework published in March was blandly permissive. Three months later Mr Trump turned draconian, banning foreigners from accessing the most powerful American AI models, prompting their maker, Anthropic, to close them to all.
More coherent policymaking is needed. Dealing with the possible long-term effects of AI, such as mass job losses or surging inequality, will require buy-in from both Congress and the president. There may be common ground. Bernie Sanders, a left-wing senator, would confiscate a 50% stake in the big AI firms and give it to ordinary Americans. Mr Trump, too, says the public should have a stake in AI firms, though he is hazier about how they would get it. A formula that gives voters a share of tech firms’ future profits would have bipartisan appeal. If they become as powerful as boosters predict, the pressure for redistribution could grow politically irresistible.
However, the details will be contentious. Many on the left, and quite a few in Silicon Valley, favour a universal basic income. Many on the right fear this would undermine people’s will to work. “I hate universal basic income,” says Mr Hawley, because “that’s just the government paying people not to work, and that ends up taking away people’s independence.”
As the midterms approach, America faces a situation with few, if any, precedents. The topic that arguably terrifies voters the most is one on which neither party has a coherent platform. Expect that to change before November. If party strategists are stumped for ideas, there’s a handy new tool that might help them.