The first wave of Anthropic’s new Public Record survey, fielded across 52,000 Americans in late 2025, reveals a public that is both eager for AI’s promised benefits and wary of its disruptions. Nearly half rank curing diseases like cancer or Alzheimer’s as a top hope, while two-thirds fear job loss. More surprisingly, this anxiety unites Americans across party lines, educational levels, and geography.
On hopes, 48% of respondents placed medical breakthroughs in their top three, followed by assistance for people with disabilities (36%) and general life convenience (23%). The demand for AI in healthcare is consistent with earlier Pew Research surveys from 2023, but the intensity here is stronger. This finding suggests that the public sees AI as a tool for solving concrete human problems, not just for commercial or entertainment purposes.
When it comes to fears, job loss dominates every state and demographic group. Sixty-four percent of Americans are worried, with Democrats (67%) and Republicans (62%) nearly identical. The fear is highest among those with postgraduate degrees—nearly ten points higher than among those with only a high school education—and lowest among daily AI users (54% versus 70% for non-users). That gap hints at a paradox: hands-on experience with AI may actually reduce anxiety about displacement, perhaps because it reveals the technology’s current limitations.
The workers who worry most about job loss are the ones whose jobs already overlap with what AI is asked to do—a pattern reflected in Anthropic’s economic research.
The second most common fear, cognitive dependency (56%), appears to be largely anticipatory. Only about one in five of those worried about dependency would feel significant disruption if AI vanished tomorrow. Yet educators are 2.5 to 3 times more likely than average to report witnessing cognitive atrophy, presumably in students. This suggests that dependency may become a real issue as AI use deepens, particularly in academic settings.
Support for government intervention is strikingly bipartisan. Seventy-one percent of Americans want the government involved in regulating AI, with 79% of Democrats and 68% of Republicans in agreement. The top areas for action are privacy (56%) and child safety (52%). This consensus is rare in an era of political polarization, and it echoes historical patterns: during earlier waves of automation, the public also called for government oversight before the disruptions became widespread.
Americans trust independent experts most (43%) and AI companies least (15%) when it comes to guiding AI development.
That trust deficit is significant. Only 15% say they trust AI companies to make decisions about how the technology is developed and used—lower than their trust in the federal government (20%) or international bodies (20%). Yet the same respondents also want companies to be held legally liable for harm (47%) and to prioritize safety over growth (44%). The message is clear: the public wants strong industry accountability, not self-regulation.
Integrated users—the 6% of Americans who use AI daily for both work and personal life—offer a preview of where mainstream opinion might go. They are less worried about nearly every harm and more trusting of all institutions, including AI companies. But their appetite for government regulation remains essentially the same as the general public’s (74% versus 71%). This indicates that even heavy users do not believe the industry can be left to its own devices.
The public’s main concern is not existential risk but immediate economic disruption—a finding that should guide both policy and product priorities.
Geographically, job loss fear varies only modestly, from 71% in Iowa to 57% in Mississippi. That narrow range suggests that the fear is not tied to current economic conditions but to a widespread perception that AI will fundamentally reshape work. In contrast, concern about misinformation (52%) and surveillance (46%) shows more regional variation, likely reflecting differences in media trust and privacy experiences.
The survey also reveals what Americans will not accept. Only 27% chose slowing AI development for safety as a top-three priority, and just 17% think development should be stopped altogether. This moderation is even stronger among integrated users, who are markedly less likely to support slowing or halting AI. The public wants guardrails, not a halt.
This research builds on Anthropic’s qualitative studies of 81,000 Claude users and its ongoing Economic Index. Together, these efforts aim to track how attitudes evolve as AI capabilities advance. The next wave of the Public Record will expand outside the US, offering a global perspective on these tensions.
For companies building AI, the survey delivers a stark warning: the public holds you responsible. For policymakers, it provides a clear mandate: act on privacy, child safety, and liability while maintaining broad support. The direction AI takes should not be set only by the companies building it—a sentiment that 85% of Americans appear to share.