Americans are optimistic about scientific advances in AI, but concerned that AI may take jobs.
Like any powerful new technology, Americans understand that AI could be used for good or bad. There is optimism on some fronts — but pessimism outweighs it today.
Most Americans do worry AI models will act on their own to harm people, and even more that models will be used by people to harm others.
Yet Americans do not want to stop AI altogether, because most do think the U.S. needs to develop AI to be competitive, and must do so faster than China. They also think that it can lead to medical and scientific breakthroughs.
So it's a matter of containing and managing it.
But who should do this? Both companies and government together.
People tend not to think AI companies will manage it responsibly by themselves, which leads to a wider call for government restrictions.
One difference between the more optimistic and pessimistic people has to do with perceived economic impact: Those who think AI is poised to take jobs are more pessimistic on other aspects about it. Views don't closely follow age or partisanship.
Companies and the government share responsibility.
And more see a government role to restrict, not promote, AI development.
However, there's at least some optimism about AI making helpful discoveries in medical treatments and making tasks easier.
That, combined with the perceived need to compete, means very few think AI should be stopped completely. Rather, they say, slow down.
There continues to be wide concern AI will take jobs. This has been a long-standing concern in previous polling.
This concern is especially pronounced among people who think the government should restrict AI use, rather than promote it.
At the same time, more workers think it will make their jobs easier to do than harder, even if they tend to think it will reduce, rather than increase the number of jobs overall.
And so on balance, it's seen as creating more problems than it solves.
This CBS News/YouGov survey was conducted with a nationally representative sample of 2,051 U.S. adults interviewed between September 16-18, 2026. The sample was weighted to be representative of adults nationwide according to gender, age, race, and education, based on the U.S. Census American Community Survey and Current Population Survey, as well as 2024 presidential vote. The margin of error is ±2.5 points.