This “Pacing the Frontier” push might sound like responsible adults trying to slow things down. But when the loudest applause comes from the biggest AI labs, my first reaction isn’t relief. It’s suspicion.
Because in tech, the cleanest way to kill competition isn’t to build a better product. It’s to write the rules.
Based on what’s been shared publicly, David Sacks — a top White House AI advisor — slammed “Pacing the Frontier” and basically waved two red flags: regulatory capture and, more dramatically, the idea it could be a kind of psychological operation. Strong words. The point, though, is pretty plain: a policy initiative is being sold as safety, and he’s warning it could really be about control.
And honestly, that worry is not paranoia. It’s pattern recognition.
When large companies support regulation, it’s rarely because they suddenly fell in love with the public good. It’s usually because they can afford it. They have lawyers. They have compliance teams. They have relationships. Regulation becomes a moat, not a seatbelt.
So imagine you’re a small AI startup. You’re not trying to “move fast and break things.” You’re just trying to compete. Now imagine a new framework shows up that says advanced AI work needs approval, reporting, audits, or whatever the final shape is. The big lab shrugs and hires three more policy people. You look at your runway and realize the paperwork alone might kill you. That’s not safety. That’s a quiet way to decide who gets to play.
The part that really matters here is the incentive. If you’re an incumbent lab, you don’t need to stop progress. You just need to set the speed limit at exactly the pace you’re already going. Conveniently, that also makes it harder for anyone behind you to catch up. You get to look responsible while staying in front.
So when Sacks frames this as regulatory capture, he’s pointing at something ugly but common: the people building the most powerful systems might also end up shaping the rules that govern those systems. Not because they’re villains. Because that’s what any rational organization does when the stakes are high. They protect their position.
Now, the counterargument is real, and it’s not stupid. AI can cause real harm. Not theoretical harm. Real harm. Imagine a tool that makes it cheap to scam your parents, fake your voice, flood your business with fake reviews, or generate targeted harassment at scale. Imagine a model that helps someone plan a bad act, or makes it easier to create dangerous instructions. It’s reasonable to want guardrails before the worst versions of this get distributed.
And if you believe the risks are big enough, then “pacing” starts to sound like basic public safety. Like seatbelts, like food inspections. Slow down. Check the work. Put lines in place before it’s too late.
But here’s where I stop buying the comforting story: the people most loudly calling for “responsible pacing” are often the same people racing hardest to ship. That doesn’t automatically make them wrong. It does make them interested.
If “Pacing the Frontier” becomes a banner that big labs can rally around, you can get a world where they keep building, but newcomers get blocked. The public gets told, “Don’t worry, we’re regulating it.” Meanwhile, the regulation is written so the current leaders remain the only ones who can comply.
That’s why Sacks’ critique matters even if you dislike his politics or think he’s being dramatic. He’s poking at the difference between safety and safety theater. And that difference decides whether we end up with a healthier ecosystem or a cartel with nicer branding.
The psychological operation comment is spicy, and I’m not fully sold on that framing. But I get the underlying fear: that the public conversation can be steered. If you can make people scared enough, they’ll accept almost any restrictions, and they won’t notice who benefits. Fear is a powerful tool, especially when the technology is confusing and the harms are easy to imagine.
There’s another consequence people don’t talk about much: global competition. If the U.S. makes it painful to build new AI companies, we don’t just slow down “bad actors.” We also slow down the next wave of domestic challengers. The winners aren’t “society.” The winners are whoever already has scale — and possibly whoever is building elsewhere under different rules.
On the other hand, if we do nothing, we’re basically running a live experiment on everyone: schools, hospitals, elections, workplaces, teenagers. And the cleanup cost lands on ordinary people first, not on the executives who shipped the product.
So yes, regulate. But don’t pretend all regulation is morally equal. There’s regulation that spreads power, and regulation that concentrates it. There’s regulation that forces transparency, and regulation that creates permission slips only the rich can afford.
If “Pacing the Frontier” is real safety work, it should make the biggest labs uncomfortable too — not just their smaller competitors.
So what would you rather risk: moving too slowly and locking AI into a few entrenched giants, or moving too fast and dealing with the harms after they’re already everywhere?