AI use is surging even as people grow more wary of it

Usage is rising; trust is not

In the United States, more adults expect AI to affect them and society negatively than positively, according to Pew Research Center. Younger people are especially pessimistic.

Other measures point in the same direction. In a May Gallup poll, 71% of American adults opposed building a new AI data center in their neighborhood, compared with 53% opposed to a new nuclear power plant. A March NBC poll found AI less popular than ICE.

Yet adoption keeps growing. ChatGPT had one billion monthly users in May, according to analytics company Sensor Tower. Google DeepMind’s Gemini was close behind, with 950 million users in July.

Pew says half of American adults now use a chatbot, more than twice the share in 2023. One in four uses one daily. Across the 38 countries in the OECD, more than a third of adults say they used generative AI tools in the past three months.

It is tempting to explain the gap by assuming that people who dislike AI are different from the people who use it. I doubt that accounts for much of it: the figures make more sense if the groups overlap heavily.

Another possibility is that using AI more makes people like it less. There is at least a correlation in that direction: people in the Global North, where AI use is highest, are more pessimistic than people in the Global South, where use is lower.

But my guess is that many people object less to the technology itself than to the insistence that it must be put everywhere. Companies are asking the public to prepare for social and economic upheaval on a scale not seen for generations. That is a hard pitch to separate from the tools they want people to adopt.

This time, the exit is not the only option

The pattern has a recent precedent. Over the past 20 years, billions of people joined Facebook and Twitter even as dissatisfaction with their owners grew. People stayed with social platforms partly because leaving meant losing posts and connections. Search followed a similar path with Google.

AI may leave users with more room to push back. All 50 US states have either passed or proposed laws on AI development and deployment. More than 2,100 bills have accumulated across the country, ten times the number from three years ago.

There are also several high-quality open alternatives to products from Google, OpenAI and Anthropic. For now, that gives users more choice and puts pressure on companies.

I don’t think that makes a change of course easy for companies worth trillions of dollars. But the future is less settled than their messaging often suggests. When I asked the head of Springboards why the company was developing a new model if he disliked AI, he said large language models could no longer be avoided — but they could be made to do something else.

That “something else” is the real test: AI that is clear about what it can and cannot do, rather than presenting itself as an approaching force that will take over the world.