Why have the heads of the biggest AI companies started talking about putting the brakes on?

Just last year, the AI debate was focused largely on one question: who would be the first to release the next, more powerful model. Now, the conversation is shifting toward slowing the process down. Companies competed with record-breaking results, benchmark scores, and promises of a universal digital assistant for everyone. Today, the tone of the conversation has changed almost completely.

Increasingly, the question is no longer how to accelerate the development of artificial intelligence, but whether we are moving too fast for our own safety. And what’s particularly telling is that this concern is being voiced not by outside critics, but by the very people creating the most powerful AI systems in the world.

TABLE OF CONTENTS:

Nadella Joins the Chorus

The final straw that made this issue truly impossible to ignore was an essay by Dario Amodei, CEO of Anthropic, in which he called for slowing down the development of the most advanced artificial intelligence models. His argument is simple yet unsettling: the capabilities of these systems are advancing faster than our ability to control their behavior.

The industry’s reaction turned out to be surprisingly unanimous. Sam Altman of OpenAI supported the idea. Elon Musk of xAI joined him. And now Satya Nadella, CEO of Microsoft – a company that has invested tens of billions of dollars in OpenAI and whose business is heavily driven by demand for AI infrastructure on Azure – has joined the list as well.

Satya Nadella’s position is framed as a simple principle: if a system being developed does not benefit humanity and is not under human control, it should not be developed. However, it’s important to note that this is not a call to freeze AI development. Microsoft is proposing something closer to a formula of “controlled acceleration” – slowing down where safety is concerned, while simultaneously speeding things up where the goal is to make the benefits of the technology available to as many countries and communities as possible.

What’s Behind the Word “Slowdown”

This is where the real intrigue begins. An attentive reader will immediately notice a curious paradox: a company whose business depends directly on how quickly demand for AI computing power is growing is publicly calling for caution. Does this signal a genuine shift in priorities, or is it more of a strategic move?

The answer is probably more complicated than a simple “either-or.” On the one hand, calls for safer AI development are increasingly coming not from outside observers, but from people who work directly on these systems every day. In recent days, statements from researchers at Anthropic and OpenAI about the risks associated with so-called recursive self-improvement have attracted significant attention – a situation in which AI takes part in creating the next, even more powerful versions of itself. According to some experts, this process is advancing faster than initially expected. To an outside observer, that sounds like something out of science fiction. But it’s important to remember: these warnings are not coming from Hollywood screenwriters, but from the engineers who are actually building these systems.

On the other hand, none of these companies is actually slowing down its own development. Billions of dollars continue to flow into new data centers, Nvidia accelerator purchases, and the recruitment of leading researchers. Everyone is talking about safety – while continuing to accelerate at the same time. And that is the essence of the current moment: the race isn’t stopping; only the rhetoric surrounding it is changing.

China Sees Things Differently

Not everyone involved in the global debate views these calls for caution as a genuine concern for safety. China’s Ministry of Foreign Affairs reacted sharply to such narratives, calling them an attempt at “fearmongering” and emphasizing that AI development should be open, inclusive, and beneficial to everyone, rather than becoming a tool for confrontation.

The reason for this reaction becomes clear when you look more closely at the details of Amodei’s essay. He called not only for slowing down the development of the most advanced models, but also for preserving the technological advantage of democratic countries over China and further restricting Beijing’s access to advanced computing power. In this context, calls for a “safe pace of development” can easily be interpreted by Beijing not as concern for humanity, but as an attempt to preserve the current balance of power, in which the United States remains the leader.

Chinese media have already begun portraying some of these initiatives as another stage in the technological Cold War. This means the debate over AI safety is becoming increasingly intertwined with geopolitics – and that makes the prospect of any international consensus significantly more difficult.

Washington Is Divided, Too

The unease is also palpable on the American political scene. Some members of Congress have already called for new safeguards to be adopted without delay, including requirements to preserve a human-controlled emergency shutdown capability for AI systems and to establish a broader regulatory framework for the most powerful models.

At the same time, House Speaker Mike Johnson takes a much more cautious position. In his view, a sweeping moratorium on AI development should be avoided, as it could weaken the United States’ position in its competition with China. Safeguards are necessary, he argues, but not at the cost of losing technological leadership. This divergence in views essentially reflects the situation as a whole: everyone involved in the debate is talking about safety, but no one wants to be the first to hit the brakes out of fear of losing ground in the global race.

A Paradox Worth Noting

The most interesting aspect of this situation is just how radically the lineup of those sounding the alarm has changed. For years, warnings about the risks of artificial intelligence came primarily from philosophers, futurists, and outside observers of the industry – voices that were easy to dismiss as theoretical or alarmist. Today, those same warnings are being voiced publicly by the leaders of Microsoft, OpenAI, and Anthropic – companies that are effectively shaping the trajectory of the technology’s development.

In this situation, Nadella is trying to strike a balance between two seemingly opposing positions. On the one hand, there is support for independent expert evaluation of AI models, stronger human oversight, and a “reasonable pace” of development. On the other, there is no intention of putting Microsoft’s own development on hold: the company continues to improve its MAI models, publish codes of conduct, and expand access to its technologies.

Perhaps this very duality is the essence of the current moment. Just a year ago, discussions like these seemed like a distant theoretical topic, of interest mainly to a narrow circle of researchers. Today, they are being debated publicly by the leaders of the world’s most influential technology companies, China’s Ministry of Foreign Affairs, and members of the U.S. Congress. Perhaps this is the most convincing evidence that artificial intelligence has ceased to be just another technology sector. It has become one of the defining political, economic, and civilizational issues of our time – and how quickly or slowly this race moves forward will now be determined not only by engineering capabilities, but also by the balance of interests, trust, and power among the world’s biggest players.

When the Leaders of the Race Hit the Brakes: Why the Heads of the Biggest AI Companies Suddenly Started Talking About Slowing Down