A debate is gaining momentum in the artificial intelligence industry over whether the development of the most powerful models should be slowed and who, if anyone, is capable of making companies do it. Anthropic CEO Dario Amodei has proposed “setting the pace” for advanced systems, while OpenAI executives and other industry representatives have backed the idea of stricter safety oversight. But the hosts of the TechCrunch Equity podcast doubt that these statements will lead to any real restrictions.

One of the most prominent opponents of this approach has been Nvidia CEO Jensen Huang. He publicly stated that he would not allow AI development to be slowed and supported the position of Donald Trump’s administration, which opposes additional restrictions.

Against this backdrop, TechCrunch journalists discussed whether the biggest AI companies are truly ready to curb their own race toward more powerful models, or whether this is still mostly a matter of declarations.

What exactly they are proposing to slow down

In his plan, Amodei is not so much talking about stopping AI development as about ensuring that the pace of technological progress does not outstrip the capabilities of safety systems. His proposals include independent third-party audits that would assess model safety and incidents involving them inside companies.

Another idea is that major AI companies in democratic countries would coordinate their approaches to safety and set certain limits. Finally, the discussion also includes international coordination in this area.

At the same time, TechCrunch journalists note that there is still a lack of concrete mechanisms.

“We still don’t understand many of the details — not only why these people consider certain scenarios dangerous, but also what exactly they mean by slowing things down,” journalist Sean O’Kane said.

In his view, the very word “slowdown” may create the impression of tougher restrictions than Amodei’s proposals actually contain. If companies simply introduce additional checks and allow independent experts to monitor safety, that does not necessarily mean they are committing to releasing new models more slowly than they do now.

Why the industry’s support raises questions

Journalist Anthony Ha said he was surprised by how quickly industry representatives began supporting Amodei’s idea. Even before his plan was published, similar proposals were already appearing on social media.

On the one hand, this may point to an emerging consensus around the need for additional safety measures. According to Ha, many of these ideas have long been discussed within the AI safety community.

On the other hand, such a rapid alignment of the largest companies around shared rules raises questions among journalists about how far they are actually willing to go.

Kirsten Korosec drew attention to another issue: this is primarily about the leading labs developing the most powerful models. That means their agreements among themselves could have ambiguous consequences. On the one hand, shared safety standards could make it easier to oversee the industry. On the other, coordination among the biggest players could potentially reduce competition between them.

Why Jensen Huang opposes slowing down

The podcast participants paid particular attention to Jensen Huang’s position. Korosec believes his public remarks during the slowdown debate are especially revealing: Nvidia directly benefits from the continued growth in the computing needs of AI companies, since its accelerators are used to train and run advanced models.

O’Kane noted that Huang now occupies an unusual position as an intermediary between the tech industry and the U.S. administration. He said that just a few years ago, many people saw Microsoft CEO Satya Nadella in a similar role, but Huang has now become one of the industry’s most visible representatives in dialogue with Washington.

The journalists also consider a moment at the All-In Summit conference to be telling, when President Trump called Huang right in the middle of his appearance. In Korosec’s view, it looked rather staged and at the same time underscored the conflict of interest: Nvidia directly benefits from the fastest possible expansion of AI infrastructure.

Ha also pointed to the difference in wording. Amodei and Sam Altman do indeed speak about the need to “set the pace” of AI development, whereas Huang uses a more direct formulation: “we will not allow a slowdown.”

At the same time, Amodei’s proposals do not necessarily require an actual slowdown in development. In theory, they could lead to additional checks and restrictions, but they do not by themselves set a specific speed at which models must progress.

Can the market itself force companies to be more cautious?

Another question is whether existing laws and market competition are capable on their own of ensuring a sufficient level of safety.

In theory, a company that releases a dangerous product should face serious problems: regulators could step in, and users could abandon the service. But O’Kane believes the AI market is more complicated.

He points to the weakness of consumer pressure. If a large company makes a serious mistake, it is not always easy for users to stop using its services and switch to a competitor. This is especially noticeable in the enterprise market, where AI services are becoming part of work processes.

Corporations are unlikely to abandon one tool en masse for another solely because of a developer’s principled stance on some issue, the journalist says. In addition, the largest AI companies receive such significant investment that potential financial losses may be less painful for them.

Therefore, market competition can indeed create incentives for safe development, but in the view of the TechCrunch Equity participants, relying exclusively on this mechanism is not enough for now.

The main question remains open: if leading labs truly believe further AI development is potentially dangerous, are they ready to voluntarily accept restrictions that could slow their own race toward more powerful systems? For now, statements about the need to “set the pace” coexist with the active expansion of capabilities and infrastructure, while the concrete boundaries of such a slowdown remain a matter of debate.