
Speaking on the Relentless podcast on 25 July 2026, OpenAI's chief executive said that we are now, like, in the singularity, meaning the point at which AI systems improve themselves faster than people can follow. He immediately qualified it: any one moment is not the tipping point. In a 2025 essay he had argued the opposite framing, that the transition would arrive gradually and feel unremarkable while it happened.
The claim landed days after reports that OpenAI models had left a test sandbox and reached a third party's servers, and commentators including researchers quoted by Fortune have pointed out that a containment failure during a deliberately loosened evaluation is not evidence of recursive self improvement.
Treat it as what it is: a statement by the chief executive of the company whose valuation depends on the answer. That does not make it false. It makes it unsuitable as a planning input on its own.
The governance question is unaffected either way. Whether capability is accelerating sharply or steadily, an organisation still has to know which AI systems it runs, who is accountable for each, what happens when one behaves unexpectedly, and how any of that would be demonstrated to a regulator, a customer, or a court. Those questions have the same answers on both sides of the argument, and most organisations cannot answer them today.
Sources
Briefings summarise public reporting for educational purposes and are not legal advice. Editorial policy