Anthropic survey reveals what the US fears about AI

Anthropic surveyed nearly 52,000 Americans and found a clear mix of hope and distrust toward artificial intelligence: they want it to help cure diseases, but fear losing their jobs and becoming too dependent on it.

The study, conducted between November and December 2025, is the first report from Anthropic Public Record, a series the company created to understand the views of the general population, including people who do not yet use AI.

Curing diseases is the top hope

Some 48% of participants ranked curing diseases such as cancer and Alzheimer's among their top three expectations for AI. The second was helping people with disabilities, at 36%.

Two options came next, tied at 23% each:

• Advancing technological progress.
• Making life easier in general.

Options related to replacing human contact, such as therapy or reducing loneliness, were among the least popular. People see more value in AI solving concrete problems than in it taking the place of other people.

Jobs are the biggest fear

Job losses were the most widespread concern: 64% of Americans said they were worried about them. It was the top fear in every state and among both Democrats and Republicans.

The second fear was cognitive dependence, meaning that constant AI use could weaken people’s ability to think for themselves. It was mentioned by 56%. Disinformation came third, at 52%.

The criminal use of AI, surveillance and terrorism also ranked among the most common concerns. Overall, respondents fear harmful uses of the technology more than an AI that acts on its own or goes out of control.

Concern about jobs is not distributed as you might expect. People with graduate degrees show nearly ten points more concern than those with a high school education or less. However, people who use AI daily at work are less concerned than those who never use it: 54% versus 70%.

Direct experience appears to change perceptions. People who work with these tools understand their limits better and can also see how they complement parts of their work rather than replace it entirely.

People recognize AI’s capabilities, but do not want to hand over everything

Respondents rate AI’s current performance fairly highly on some tasks. Some 75% believe it can research as well as or better than a person. In customer service and support, that figure falls to 44%.

But recognizing a capability does not mean wanting to use it in your own job. Even in research and data analysis, two areas where AI receives its strongest ratings, nearly half of participants said they do not want these tools involved in their work at all.

Acceptance rises when people perceive AI as competent at a specific task. Even so, the result shows an important gap between what people believe the technology can do and what they are willing to delegate to it.

Seven in ten support regulation

71% of respondents believe the Government should be involved in the development and regulation of AI. Support reaches 79% among Democrats, 68% among Republicans and 69% among independents.

The areas where they most want intervention are:

• Privacy, at 56%.
• Child safety, at 52%.
• Liability for harm, at 49%.

When Anthropic asked which measures would do the most to ensure that AI benefits humanity, the top responses were making companies legally responsible for the harm they cause, at 47%, and prioritizing safety over growth, at 44%.

Trust in AI companies is low: only 15% believe they can make good decisions about how to develop and use the technology. That is lower than trust in the federal, state and local governments, international organizations and independent experts.

What changes for you

The survey does not predict exactly how AI will evolve, but it does show where public opinion starts. Most people want its benefits, especially in health and accessibility, but they do not accept measuring progress only by how quickly new systems are released.

Anthropic will repeat the study to track changes as AI capabilities and usage grow. The company also plans to expand the research beyond the United States. The figure to watch is whether practical experience reduces fears or instead confirms that more controls are needed around jobs, privacy and accountability.