Let it rip. These three simple words have long summed up the Trump administration’s approach to AI regulation. When Donald Trump returned to the White House, his team swept away what they saw as the nannying, safety-obsessed overkill of the Biden administration. At an international AI-safety summit, his vice-president, J.D. Vance, was blunt: “Excessive regulation…could kill a transformative industry just as it’s taking off.”

Roughly two years later, AI has certainly taken off. The models’ capabilities are improving at a jaw-dropping pace. Fear is growing, too, reflected by the steady stream of existential warnings from inside AI giants such as OpenAI and Anthropic. And data centres have become a politically toxic focus of ordinary Americans’ AI concerns. Yet letting it rip is still the administration’s official position. Mr Trump has rejected calls for regulation of what he now insists on calling “Super Intelligence”. He has focused on the importance of America winning the geopolitical race with China; and late last month struck an agreement on safety with the AI industry that is only “morally binding”, ie, not especially binding at all.

Can this laissez-faire approach last? I’ve just returned from a ten-day trip to America where I spent much of my time talking to tech and business leaders, financiers and Washington insiders about whether, when and how America will take a firmer regulatory stance towards AI. And in this week’s episode of The Insider, I’ll be discussing that question with my colleagues. I’ll be joined by Edward Carr, tomorrow’s host and deputy editor; Charlotte Howard, our US editor; and Alex Hern, our lead AI writer. Shashank Joshi, our Washington bureau chief, will join remotely, too.

First, we’ll examine the dilemma. Mr Trump is stuck between a bot and a hard place. If he continues a light-touch regulatory regime, his unpopularity could surge even more. But if he decides to regulate, he could crash the stock market and set America behind China in the race for AI supremacy.

More than 90% of Americans want guardrails on AI. Most see some risk that AI could destroy humanity. Charlotte and Shashank will offer insights into this growing angst. The risks from lax rules go further than electoral politics. The Hugging Face incident, in which a “swarm” of rogue agents attacked the AI platform, hinted at what could lie ahead. Alex will explain if we can expect such dangerous cyber-attacks to proliferate.

But what might happen if Mr Trump were to suddenly tighten the leash? The accelerationists advising the president argue that the existential risks are exaggerated and existing rules are sufficient. People like David Sacks, a venture capitalist, warn that hasty regulation will bring an end to America’s AI boom. That would have a disastrous effect on the valuations of the companies that this week drove American markets to an all-time high—something Mr Trump cares about deeply. He would not want to be the president who gave China the lead in the century’s most important technology.

We will discuss how much truth there is to these claims. I certainly think there is some risk of regulatory overkill. Governments are notoriously inept at getting regulation right, particularly of a technology that so few people understand. And the politics will only get more difficult. I’m keen to hear from my colleagues about what it would take for Mr Trump to change tack. Will he be influenced by growing Republican opposition to AI? And what effect might a drubbing in the midterms have?

But, even if the politics permit it, what might better regulation look like? Can existing criminal and civil law really be applied to something as complex—and fast-moving—as AI? And how do we make sure that models are properly vetted, evaluated and tested?

I expect we will disagree on the perfect approach. We want to hear your views, too. Vote in our poll and write to us at insider@economist.com.

The Insider will be available to watch from 6pm London time (1pm in New York) on Thursday. Please tell us what you want to know by using the Q&A feature on the episode page—we’ll do our best to answer your questions. See you tomorrow.