Picture a Vatican diplomat trying to deliver a moral warning about artificial intelligence while a roomful of lobbyists clinks forks against salad plates and quietly trades business cards. That happened. Last week, at a black-tie dinner inside the Waldorf Astoria in Washington, the hotel that once carried Donald Trump's name.
Archbishop Gabriele Caccia, the Holy See's top envoy to the United States, showed up unannounced. His job was to relay a message from Pope Leo XIV: protect people and human dignity before chasing profit and the next breakthrough. The crowd, gathered for an awards gala thrown by the Washington AI Network, was there for something else. Networking. Trophies. A reminder of who matters.
The nuncio, by one account from The Verge's Tina Nguyen, was nearly inaudible over the cutlery.
That scene tells you almost everything about how AI regulation actually works in the capital right now. The pope wrote a whole encyclical, Magnifica Humanitas, laying out Catholic doctrine on the technology. People outside Washington seem genuinely moved by it. Inside the room? Polite nods, then back to the steak.
The reason is blunt. The pope can't write a statute or hand down a fine. No legal teeth, no urgency. In this town, influence is measured in enforcement power, and the Vatican has none of it.
The strangest guest list in tech policy
The attendee mix alone is worth a pause. AI lobbyists. Safety nonprofits that spend their days warning about the same products those lobbyists are pushing. Tech reps. Journalists. Senior officials, including Mehmet Oz, now running the Centers for Medicare and Medicaid Services, and Darío Gil, an under secretary at the Department of Energy. Dancers on stilts dressed as robots. And Kevin O'Leary, the Shark Tank investor, who walked off with an award.
Let that last one sit for a second. The celebration of AI's breakout figures was broad enough to honor a reality-TV financier. It's the kind of detail that makes you wonder what the bar for "power player" even is anymore.
The whole event was a snapshot of an industry with tunnel vision. Everyone in that room is chasing the same prize, and most of them are anxious about not getting it. Hence the strange bedfellows. A church official, a TV personality, federal regulators and the people trying to regulate them, all in one ballroom, all circling the same question: who gets to decide how AI is built and sold.
Why the old lobbying playbook broke
For decades, corporate lobbyists ran a simple strategy. Make friends on both sides. Donate to Democrats and Republicans, keep every door open, never burn a relationship you might need in six years when the majority flips. Patience was the whole game.
That playbook doesn't work in Trump's Washington. Backing a Democrat at any point in your past can now read as betrayal, even if you're a tech billionaire with more money than several small nations. The example everyone keeps citing is Jared Isaacman, the entrepreneur and private astronaut Trump tapped to lead NASA. His nomination reportedly stalled for months once the president learned Isaacman had once cut a check to a Democrat. One donation. Months of limbo.
So the math has shifted. If you want regulatory favors, you don't spread your bets. You pick a side, write the checks, and make the president look good. Do that, and Trump can be talked into delivering more or less any rule change you ask for, then leaning on congressional Republicans to fall in line behind him.
The catch is that loyalty buys you less certainty than you'd hope. Trump's attention is a moving target, and the people closest to him are forever pulling it in different directions.
A single executive order, three different decisions
Want proof of how unstable this all is? Follow one document over about two weeks.
On May 20th, multiple outlets reported that the president was about to sign an executive order creating a federal review board for advanced AI models that hadn't yet been released. The board could delay a launch by up to 90 days while it took a look.
Then on May 21st, Trump backed off. Didn't sign it. The reporting points to David Sacks and Elon Musk, who reportedly leaned on him at the eleventh hour and got him to walk it back.
Then he changed his mind again. On June 2nd he signed the order after all, reportedly after Treasury Secretary Scott Bessent made a fresh push. But the terms had moved. The maximum delay dropped from 90 days to 30.
Three positions, two weeks, a handful of billionaires whispering in his ear. That's not a process. It's a tug-of-war, and the rope is the future of AI regulation in the United States. Anyone trying to build a company or a compliance plan around it is essentially guessing.
Here's the part that should worry people who don't follow Washington for a living. The rules governing some of the most consequential technology of the decade are being set, unset, and reset based on who got the last meeting. Not hearings. Not careful study. Whoever caught the president's ear most recently.
The two dates everyone can actually plan around
There is one bit of predictability in the whole mess. However erratic things get, two events arrive on schedule. Every two years, a federal election lands in November. The winners take their seats in Congress the following January. The balance of power shifts, reliably, on that clock.
What nobody can predict is the direction. Who controls the House and Senate after the 2026 midterms, what they believe about AI, whether they want guardrails or a clear runway for the industry, it's all up for grabs. For tech companies, that uncertainty stacks on top of the day-to-day unpredictability of the White House. An endless line of unknowns sitting underneath every long-term call they make.
Which is why the gala matters more as a symbol than the encyclical does. The pope offered a moral framework with no deadline. The midterms offer a hard deadline with no clear outcome. Washington's AI crowd is watching the second one, because it can actually move the levers they care about.
Still, it's hard not to find something a little melancholy in that ballroom. A man from the Vatican stood up to argue that human beings should come before profit, and the dominant sound in the room was forks. The message wasn't rejected so much as ignored. Filed under "nice, but not enforceable."
The real fight over AI regulation won't be settled by an encyclical, or even by an awards dinner. It'll be settled by which lobbyist gets the next meeting, and by who's sworn into Congress next January. Watch those races. The companies in that room already are.
And if you're wondering whether three contradictory decisions on one executive order is a temporary glitch or the new normal, the honest answer is that nobody knows. Including, probably, the people making the decisions.