Most Americans Think AI Will Harm Them. Here Is What the Evidence Says

Three separate polls conducted in early 2026 reached the same conclusion: a majority of Americans believe AI poses more risks than benefits. An NBC News poll of 1,000 registered voters put that figure at 57%. A Quinnipiac survey found 55% expect AI to do more harm than good in their daily lives, an 11-point rise from a year earlier. Pew Research found most Americans are more concerned than excited about AI's spread. These are consistent results from credible pollsters, not outliers.

The question worth asking is whether those concerns are grounded in evidence, or whether they reflect something else.

What people are actually worried about

The NBC poll found that 70% of Americans believe AI will reduce job opportunities. Voters identified workers and the environment as the two areas most likely to be harmed by AI. Trust is a factor too: 76% of respondents say they trust AI "hardly ever" or "some of the time."

The demographic breakdown is worth noting. Adults aged 18-34 have a net AI favorability rating of minus 44. Women aged 18-49 sit at minus 41. Men over 50 and upper-income voters are the only groups with net positive views. AI is perceived as a threat primarily by the people who tend to be in more economically precarious positions. That may not be irrational.

The expert-public gap

The Stanford AI Index 2026 documented a striking divergence: 73% of U.S. AI experts view AI's impact on jobs positively, while only 23% of the general public agrees. A 50-percentage-point gap. Some of that reflects genuine information asymmetry — experts know more about how AI actually works and what it can and cannot do. But some of it reflects a less flattering truth: experts in AI tend to work in fields where AI is a tool for their use, not a threat to their livelihood.

Globally, the picture differs. Ipsos data from the same report found 59% of people worldwide believe AI offers more benefits than drawbacks, up from 55% in 2024. American skepticism is not the global norm.

Are the fears warranted?

On jobs: the evidence is genuinely mixed. Some roles are being displaced; others are being augmented. Productivity gains from AI have been real, but whether those gains translate into broadly shared prosperity or concentrate among shareholders and top earners is an economic and political question as much as a technological one. The 70% who believe AI will reduce job opportunities are not obviously wrong. They may simply be focusing on disruption costs while underweighting adaptation and new job creation. Both things can happen at once.

On trust: the lack of transparency in how many AI systems make decisions is a legitimate concern, not a misconception. Systems that affect hiring, credit scoring, and medical triage are frequently opaque. Skepticism about trusting those systems is reasonable, and researchers building them largely agree.

The honest read is that public fears are a mix: some well-founded concerns, some reasonable uncertainty, and some misconceptions amplified by media coverage that gravitates toward alarming scenarios. The expert consensus that AI is net-positive in aggregate is probably correct over the long run. But "aggregate" and "long run" offer cold comfort to someone in a job being automated right now. The gap between what experts believe and what the public feels may tell us less about who is right and more about who bears the cost.