Anthropic publishes the first results of its Anthropic Public Record, a national survey conducted at the end of 2025 that collected the opinions of 51,993 Americans. What do people expect and fear about artificial intelligence? Here’s the picture, clear and without technical jargon.
What Anthropic measured and why
The survey aimed to capture the general public’s attitude —not just AI users— about benefits, risks and governance. It was conducted online by YouGov between November and December 2025 and weighted to be representative by age, gender, education and ethnicity.
Why does this matter to you? Because understanding what everyday people think helps design policies, products and rules that actually respond to their expectations and fears.
Main findings
-
Hopes: almost half (48%) listed curing diseases as one of their top three expectations for AI. Helping people with disabilities was second at 36%. Technological advances and making daily life easier tied at 23%.
-
Fears: the top-cited issue was job loss due to AI, with 64% worried. That’s followed by cognitive dependence (56%) and misinformation (52%).
-
Government and intervention: more than 70% believe the government should play a regulatory role in AI; that support was bipartisan. Priorities: privacy (56%), child safety (52%) and liability for harms (49%).
-
Trust in industry: only 15% trust AI companies to decide how the technology is developed and used.
Overall, the survey shows a consensus: people want the benefits of AI, fear its disruption and demand accountability.
Details that help understand the numbers
-
Concern about job loss is widespread: it’s the top worry in every state and among voters of both parties.
-
Interestingly, worry about job loss rises with higher education: those with postgraduate degrees are nearly 10 percentage points more concerned than those with a high school education or less. This suggests that jobs closer to tasks AI can do are the ones that worry people most.
-
People who use AI every day at work are less fearful: 54% of them worry about job loss versus 70% among those who don’t use AI. Does using AI boost confidence, or does experience reveal real limits? Probably a bit of both.
Cognitive dependence: anticipation more than reality, for now
Cognitive dependence was the second most frequent fear, but evidence suggests much of it is anticipatory. Of those who said they worry about dependence, only about a fifth reported they would suffer major disruption if AI stopped working tomorrow.
Also, teachers report more signs of cognitive atrophy among students, a finding that aligns with Anthropic’s qualitative research with Claude users.
The "integrated users": a glimpse of the future
Around 6% of respondents used AI daily both at work and in personal life. This group is younger, more urban, more educated and more inclined to experiment with tech.
Integrated users tend to trust institutions more and are less in favor of slowing AI development. Still, they support government intervention at rates similar to the general public.
What people ask for in terms of governance
When asked what would make AI benefit humanity, the top responses were:
- Making companies legally responsible for harms (47% put it in their top 3).
- Prioritizing safety over growth (44%).
There was also support for independent bodies with real power and for measures to slow development if necessary.
Methodology in brief
- Sample: online representative survey of 51,993 U.S. residents, ages 16+.
- Period: November 1 to December 11, 2025.
- Sampling and weighting: YouGov panel, adjusted to census benchmarks.
- National margin of error: ±0.6 percentage points at 95% confidence.
What Anthropic will do with this and what’s next
Anthropic will use this survey —and other tools like large-scale interviews and usage data— to refine policies and products. Their public proposals include frameworks for independent safety testing, transparency requirements and economic policy tools to mitigate labor impacts.
The survey will be repeated regularly and is planned to expand beyond the United States. It’s a way to measure how opinions change as the technology advances and adoption grows.
Practical takeaway: people want progress, but not at any cost. They ask for concrete benefits like better health and support for people with disabilities, and at the same time demand clear rules and responsibilities. If you work in tech, policy or education, this is the map of expectations you should be working with.