The latest count comes from the United States. In an AP-NORC poll published on Thursday 8 October, 64 percent of American adults said AI is developing too fast. Only 8 percent said too slow. About eight in ten said it is extremely or very important that the government keeps AI under human control. The worry crosses party lines: about three quarters of Democrats and around half of Republicans said too fast.
It is not only an American mood. Pew Research asked people in 25 countries last year whether AI makes them more excited or more concerned. In no country did more than 30 percent say they were mainly excited. Ipsos, polling 32 countries this spring, found wonder and worry almost evenly split, with many people feeling both at once.
What is new is who joined the chorus. On 12 September, Anthropic’s chief executive Dario Amodei published an essay calling on labs to pace the frontier, so that safety research and outside checks can catch up. Within days the heads of OpenAI, Google DeepMind, Microsoft and xAI voiced support, each in their own words. Meta’s Mark Zuckerberg disagreed, and President Trump dismissed calls to slow down, saying America must stay ahead of China.
So the people, the pollsters and most of the builders agree. And yet no lab has announced a pause or withdrawn a product. Only Anthropic made a concrete, checkable promise: permanent access for outside evaluators. The others have promised to talk about standards.
That gap between saying and doing is the story. Everyone agrees on the speed limit; nobody has touched the brake. For now, the most honest measure of AI’s pace may be how quickly “we should slow down” has become something everyone can say without having to do it.