A large majority of Americans are worried that artificial intelligence could permanently displace workers, according to a new Reuters/Ipsos poll released this week.

The six-day survey, which concluded on Monday, found that 71% of respondents fear AI will “put too many people out of work permanently.” The findings highlight growing unease over the technology’s impact, even as the U.S. unemployment rate stood at a relatively low 4.2% in July.

Since OpenAI’s launch of ChatGPT in late 2022, artificial intelligence has surged into public awareness, spurring rival products from Meta, Alphabet and Microsoft. While AI adoption has driven investment and innovation, it has also raised alarm over risks to employment, politics, national security and social life.

According to the poll, 77% of Americans worry AI could be used to fuel political unrest, reflecting concerns over deepfake technology. The issue drew renewed attention last month after former President Donald Trump posted an AI-generated video falsely depicting Barack Obama being arrested.

Military applications of AI also divide public opinion. Nearly half of respondents (48%) said the U.S. government should never use AI to select military strike targets, compared with 24% who said it should be allowed.

Beyond security concerns, 61% of those surveyed expressed unease about the large amounts of electricity required to power AI systems. Companies are responding: Google recently struck deals with two utilities to curb its data center energy use during peak demand periods.

The technology’s social impact is also under scrutiny. Two-thirds of Americans worry people may abandon human relationships in favor of AI companions. Respondents were split on whether AI will improve education, with 36% optimistic and 40% skeptical.

The Reuters/Ipsos poll surveyed 4,446 U.S. adults nationwide online and carries a margin of error of about two percentage points.