“Looking at what’s possible, it does feel surprisingly slow,” Sam Altman said recently on the topic of AI adoption rates. It is no surprise to see technology leaders bemoan the pace of AI diffusion – they have something to sell, after all. But why do so many political leaders echo their impatience for greater speed?

I do understand the underlying rationale. If AI can boost productivity growth – and I think it might well do in a number of sectors – then the sooner you can spread it, the sooner you can reap the benefits for growth. For the UK government in particular, which is trying to jolt into life a moribund economy, the appeal is understandable.

But bullish rhetoric about “unleashing” AI is badly out of touch with the sensibilities of many voters in the developed world. A poll in the US last month found that just 6 per cent of registered voters thought the use of AI was moving too slowly, while 30 per cent thought it was moving at the right pace, and 60 per cent felt it was moving too quickly.

Leaders like Donald Trump often talk about an AI “race”, but I suspect that many people either don’t know what the finish line is supposed to be, or have only heard it articulated by technology executives, whose predictions of the future don’t exactly sound like a place that most normal people would want to accelerate towards. Anthropic’s Dario Amodei foresees “unusually painful” labour market disruption at a pace that is “hard for people to adapt to”, while Altman recently said: “We see a future where intelligence is a utility like electricity or water and people buy it from us on a meter.”

Of late, politicians have moderated their tone somewhat. Trump has told AI companies they need “PR help” while UK chancellor of the exchequer Rachel Reeves acknowledged people’s concerns in a recent speech. “There will be bumps in the road as we navigate these changes,” she admitted, while saying she wanted Britain to “achieve the fastest rate of AI adoption” in the G7. But the attempt to combine both messages doesn’t work well. Most people know that bumps in the road ahead aren’t usually a cue to press the accelerator.

If the politics are tricky, the economic case for hurrying this transition along isn’t completely clear-cut either. We know from history that people and economies are remarkably good at adapting to technological change over the long run. In 1920, more than 14 per cent of people worked in agriculture and mining in the UK, for example. By 2016, that share had fallen to almost 1 per cent. In many professions and sectors that have declined gradually over time, older people have retired and younger people have not joined.

But economic change can be painful if it outpaces people’s ability to adapt. MIT professor David Autor’s work on the “China shock”, for example, highlighted the impact of Chinese competition on American textile mill workers, many of whom were geographically concentrated and struggled to get out of harm’s way fast enough.

A new paper by Argentine economist Eduardo Levy Yeyati argues that two economies can arrive at the same technological frontier, and yet experience different transitions. “In one, firms adopt AI at a pace the labour market can absorb,” he writes. “In the other, adoption outruns reallocation capacity. The long-run destination may be identical; the permanent social damage is not.”

The public is aware, too, that general purpose technologies always lead to social and cultural as well as economic change. Indeed, when Anthropic recently interviewed 80,000 users of its Claude chatbot across 159 countries, many of the concerns people raised were outside the economic realm: autonomy and agency; cognitive atrophy; misinformation; privacy; wellbeing and dependency.

The economic historian Karl Polanyi once argued that “it should need no elaboration that a process of undirected change, the pace of which is deemed too fast, should be slowed down, if possible, so as to safeguard the welfare of the community”.

This does not have to be an argument against change – even rapid change. The key phrases, for me, are “undirected” and “deemed too fast”. Whether a process feels “too fast” depends on the psychological and practical capacity of society to cope with the pace. That capacity can be improved via better safety nets and proactive skills policies that enable people to pre-empt and drive change, rather than merely react to it.

But at the moment, I think many people feel they are in a speeding car without either a seat belt or their hands on the wheel. In those circumstances, nobody should be surprised if they vote to hit the brakes. – Copyright The Financial Times Limited 2026