Should AI Slow Down, or Is This Just Politics?

Anthropic CEO Dario Amodei called for the AI industry to slow down. Sam Altman and Elon Musk agreed. Days later, Donald Trump argued for the opposite. So what are we actually looking at: a real AI safety problem, or another fight over who gets to control the technology?

On September 12, Dario Amodei, CEO of Anthropic, published a lengthy essay arguing that artificial intelligence companies should slow down the pace at which they improve their models.

The proposal was not simply to "stop AI."

Amodei argued that the pace of development should be accompanied by stronger safety testing, independent evaluations, and greater coordination among companies developing the most advanced models. His concern is that AI capabilities may be advancing faster than our ability to understand and control their consequences.

That alone would not be particularly surprising coming from the CEO of a company that has built much of its identity around AI safety.

What made the story interesting was who agreed with him.

When competitors agree

Shortly after Amodei's essay was published, OpenAI CEO Sam Altman said he agreed that the development of frontier AI models needed to be paced.

Elon Musk also publicly supported Amodei's position.

Demis Hassabis of Google DeepMind expressed support as well. Suddenly, some of the most influential people in AI were making an argument that would have been difficult to imagine a few years ago:

perhaps the race toward the next generation of AI needs some kind of brake.

There is an important detail here.

They were not calling for AI development to end.

The debate is about pace.

The idea is to keep moving forward while making sure that safety testing, evaluation, and oversight can keep up with the technology.

Then Trump entered the discussion

A few days later, the debate took on a completely different dimension.

Donald Trump pushed back against the idea of slowing the AI race. The central argument was straightforward: the United States cannot slow down while it is competing with China for technological leadership.

From the administration's perspective, slowing down unilaterally could give China an opportunity to move ahead in a technology increasingly viewed as strategically important to the economy and national security.

And this is where the conversation stops being only about safety.

It becomes a question of power.

So who is right?

That may be the hardest question to answer.

There are legitimate safety concerns.

AI models are becoming more capable, performing increasingly complex tasks and taking part in processes that previously required constant human intervention.

At the same time, there is an enormous economic competition between the companies building them.

OpenAI, Anthropic, Google, xAI and others are competing for users, developers, enterprise customers and infrastructure.

And there is also competition between countries.

The United States does not want to lose technological leadership to China.

So we end up with a rather strange situation:

companies can genuinely believe that AI needs stronger safety measures while also not wanting to be the first company to slow down.

The "first to slow down" dilemma

Imagine two companies developing increasingly powerful AI models.

The first decides to slow down and spend more time on safety testing.

The second keeps moving.

The first company could lose market share, talent and technological advantage.

Now apply the same logic to two countries.

If the United States slows down while China continues accelerating, Washington could see that as a strategic disadvantage.

And the same logic applies in the other direction.

That is what makes global coordination so difficult.

It is not enough for everyone to agree that risks exist.

Everyone also has to believe that the others will follow the same rules.

There is another interesting detail

While some of the biggest names in the industry are calling for caution, their companies are still competing to release better models.

Anthropic, for example, is considering new model releases at a time when it is facing increasing competitive pressure from OpenAI.

That does not necessarily mean there is a contradiction.

Safety and development can happen at the same time.

But it does show how difficult a real "pause" would be in an industry where billions of dollars are being invested and where every major breakthrough can create a significant competitive advantage.

So is this all politics?

Probably not.

It would be too simplistic to say that AI safety warnings are merely a political strategy designed to control the market.

There are real risks, and researchers and companies are genuinely trying to find ways to evaluate them.

OpenAI, for example, has been advocating for international technical standards, common evaluation methods and incident-reporting mechanisms for advanced AI systems. The company also argues that this is not about stopping progress, but about creating the conditions for safer progress.

But it would also be naive to ignore the political dimension.

AI has become a strategic issue for both the United States and China.

Chips, data centers, energy, AI models, intellectual property and computing capacity are increasingly part of the same competition.

Safety and geopolitics can therefore be happening at the same time.

Maybe the question is wrong

Perhaps the question isn't simply:

"Should AI slow down?"

Maybe the more interesting question is:

"How much can we slow down without losing the race?"

And that question has no simple answer.

If the risks are genuinely significant, moving faster without understanding the consequences could be a terrible strategy.

On the other hand, if one country or company slows down while its competitors continue advancing, it could give away exactly the advantage it was trying to protect.

It is a paradox.

The more everyone fears that someone else will move ahead, the harder it becomes for anyone to stop.

In the end, it may be a mixture of both

It is difficult to look at this debate and conclude that it is all just political theater.

But it is equally difficult to ignore the technological race taking place.

Amodei may genuinely believe that AI models are advancing too quickly.

Altman may genuinely agree with him.

Musk may genuinely believe that the risks need to be addressed.

And Trump may genuinely believe that slowing down the United States would hand an advantage to China.

Those positions are not necessarily incompatible.

That may actually be the most interesting part of the entire debate.

The AI race has reached a point where safety, money, technology and geopolitics are becoming deeply intertwined.

And perhaps nobody wants to be the first one to hit the brakes.