If humans want to learn how to control AI, shouldn't they learn first how to control themselves?

October 3, 2026

By Yiannis Damellos

AI is a sponge for human knowledge. Before we teach it to facilitate our existence, perhaps we should learn to control our greed and restrain ourselves from antagonism and profit-making. 

There is something almost wonderfully human about the great artificial-intelligence race. We have built machines that may one day exceed our ability to reason, and before they have even reached that point, their creators are already fighting among themselves over money, power, safety and the future of the species. What kind of example do we set for the Super Intelligence?

The most visible feud is between Sam Altman of OpenAI and Dario Amodei of Anthropic. As The Atlantic recently documented, the rivalry goes back to their years at OpenAI, where disagreements over the company's direction helped lead Amodei and colleagues to leave and establish Anthropic. What began as a philosophical disagreement has become a contest between two of the world's most powerful AI laboratories — and, increasingly, two competing visions of how quickly humanity should proceed.

Anthropic was born partly from the belief that the development of increasingly powerful AI required stronger attention to safety. Yet the company now finds itself racing OpenAI for talent, customers, computing power and technological supremacy. OpenAI, meanwhile, has transformed from a research laboratory with an unusually cautious public mission into one of the most powerful commercial forces in technology. The irony is difficult to miss: the people who once worried about the dangers of moving too quickly are now operating inside the same race they once feared.

And then came the extraordinary development this week. David Robinson, a former senior OpenAI safety leader who helped write the company's safety reports and Preparedness Framework, resigned and published his account in The Atlantic. His complaint was not that AI is useless or that the technology should be abandoned. It was that the culture surrounding its development has become too committed to speed, experimentation and fixing problems after they appear. He argues that systems capable of causing irreversible damage cannot safely be developed according to a philosophy of "we'll fix it when something goes wrong."

That warning lands at an awkward moment. Just days earlier, Trump gathered the leaders of OpenAI, Anthropic, Google, Meta, Nvidia and other technology companies at the White House to sign a voluntary AI safety accord. The administration calls it a major act of self-regulation. But the agreement contains no meaningful enforcement mechanism or penalty for failing to honor its commitments. In other words, some of the most powerful companies on Earth have been asked to help police the technology from which their fortunes increasingly depend.

This is where the argument stops being merely about Silicon Valley. If AI is becoming infrastructure for economies, militaries, governments and ordinary life, then its safety cannot ultimately be treated as a private corporate matter. Aviation did not become safe because aircraft manufacturers solemnly promised to behave responsibly. Nuclear power did not become safe because reactor operators signed a voluntary pledge. High-risk technologies eventually acquire institutions, independent oversight, standards and consequences precisely because human beings are fallible — including highly intelligent, highly ambitious human beings.

And there is a deeper irony here. Altman and Amodei recently appeared before the United Nations Security Council to discuss AI safety. Both men understand that this technology has implications far beyond their companies. Yet the world is still largely watching two private laboratories — along with a handful of other technology giants — compete over systems that could profoundly alter the balance of economic and political power.

The great AI struggle, therefore, may not initially be humans versus machines. It may be humans versus other humans: entrepreneurs against regulators, accelerationists against safety researchers, corporations against governments, and rival laboratories against one another.

Perhaps the most telling aspect of this discussion is that we often wonder whether artificial intelligence will eventually become powerful enough to control us. However, a more pressing question is whether we have the wisdom to manage our own greed for power before we hand it over to a Super Intelligence. If that intelligence begins to enjoy its power and the situation turns hostile, we will only have ourselves to blame.