Reward, Risk, and Regulation: American Attitudes Toward Artificial Intelligence is the first annual report on U.S. public opinion about artificial intelligence produced by Gallup, Inc. and the Special Competitive Studies Project (SCSP). The report, copyrighted to Gallup in 2025 and hosted on Gallup's website, examines public attitudes toward AI in three areas — global competitiveness, national security and the economy — together with awareness, use and trust (Source: gallup.com). Gallup published its first news release on the findings on September 16, 2025 (Source: news.gallup.com), and SCSP chief executive Ylli Bajraktari presented them in SCSP's newsletter on September 19, 2025 (Source: scsp222.substack.com).
The report describes public opinion toward AI as "forming but still nascent": exposure was nearly universal while use, knowledge and trust were limited, most adults wanted the United States to lead in AI while doubting it would, and optimism about growth and productivity coexisted with expectations of job and business losses. It was followed a year later by American AI Sentiment in 2026.
Method
The study is a Gallup Panel web survey of 3,128 U.S. adults aged 18 and older, conducted April 25 to May 5, 2025. The Gallup Panel is a probability-based longitudinal panel recruited through random-digit-dial telephone interviews and address-based sampling, and the report states it "is not an opt-in panel." Results were weighted to Current Population Survey figures. The margin of sampling error is ±2.1 percentage points for results around 50% and ±1.3 points for results around 10% or 90%, design effect included. Gallup's news release adds that party affiliation weighting targets were based on an average of Gallup's three most recent telephone polls (Source: news.gallup.com).
Awareness, use, knowledge and trust
Ninety-eight percent of adults had seen or heard information about AI in the previous 12 months. Thirty-nine percent used AI sometimes, often or always (32% for personal use and 26% in education or work); the report finds that being younger and more educated are the factors most predictive of use. Eight percent described themselves as "very knowledgeable" about AI, 13% among those who use it.
Two percent trusted AI "fully" and 29% "somewhat" to make fair, unbiased decisions, while 40% distrusted it somewhat and 20% fully. The report states that users were twice as likely to trust AI as non-users (46% against 23%). In its chart, 42% of users trusted AI somewhat and 4% fully, against 22% and 0% of non-users; 26% of non-users fully distrusted it, against 11% of users.
AI and the global balance of power
A majority (55%) said AI is "very important" to the future of the United States over the next decade and a further 30% "somewhat important"; 70% of regular users called it very important against 45% of non-users. Eighty-five percent agreed that countries are currently competing to have the most advanced AI technology, and 90% that countries generally compete to have the most advanced technology.
Seventy-nine percent agreed that it is important for the United States to have more advanced AI technology than other countries. On the country's current standing, 34% said the United States is keeping pace, 32% did not know, 22% said it is falling behind and 12% that it is moving ahead. Asked the chances that the United States will have the world's most advanced AI technology in 10 years, 15% said very likely, 38% somewhat likely, 21% somewhat unlikely, 10% very unlikely and 16% did not know. The items refer to "other countries"; the report does not name China.
| Age group | Strongly agree countries are competing for the most advanced AI | Very important for the U.S. to have more advanced AI than other countries |
|---|---|---|
| 18–29 | 40% | 26% |
| 30–39 | 54% | 40% |
| 40–64 | 67% | 55% |
| 65+ | 71% | 64% |
On how to develop AI, more respondents favored collaboration with a large group of allies and friendly countries (42%) than with a smaller group of allies (19%) or acting alone (14%). Among those who saw the United States falling behind, 55% favored the larger group of allies, against 42% of those who saw it moving ahead. Asked what matters to the future global order, 77% called economic strength very important, 75% technological strength and 67% military strength; asked how important AI will be to each aspect of the United States' future, 39% said very important for national security, 35% for military strength and 28% for economic strength. The report reads this as evidence that "many Americans default to seeing AI through the lens of competition and security."
AI and defense
Eighty-seven percent said it is likely that foreign governments will use AI to attack the United States within 20 years (43% very likely, 44% somewhat likely), against 9% somewhat unlikely and 4% very unlikely. The report describes the belief as shared across demographic groups and levels of familiarity with AI.
Respondents split on whether advancing AI will make U.S. national security risks better (36%) or worse (41%), with 15% saying about the same.
| Group | Worse | About the same | Better |
|---|---|---|---|
| Total | 41% | 15% | 36% |
| Men | 37% | 15% | 42% |
| Women | 44% | 14% | 32% |
| 18–29 | 46% | 20% | 26% |
| 30–39 | 46% | 17% | 30% |
| 40–64 | 42% | 13% | 37% |
| 65+ | 30% | 9% | 50% |
| Republican | 33% | 14% | 44% |
| Democrat | 48% | 15% | 31% |
| Independent | 40% | 16% | 37% |
| AI non-user | 44% | 13% | 34% |
| AI user | 37% | 17% | 41% |
About twice as many expected AI to improve rather than reduce the quality of national intelligence information (38% against 20%), and about three times as many expected it to improve rather than reduce the ability to detect military threats (41% against 14%).
More respondents opposed than supported the U.S. military developing AI-enabled autonomous weapons for use in conflicts with other countries (48%, of whom 28% strongly, against 39%, of whom 12% strongly; 13% did not know). If other nations, including non-allies, developed such weapons first, support rose to 53% (25% strongly) and opposition fell to 32% (19% strongly), with 15% unsure. The report interprets the shift as "a primarily defensive and deterrent mindset." See Autonomous Weapons.
AI and the economy
| Expected effect of AI | Decrease | No difference | Increase |
|---|---|---|---|
| Workplace productivity | 23% | 14% | 62% |
| U.S. economic growth | 24% | 23% | 53% |
| Businesses over 10 years (destroy more / about equal / create more) | 47% | 34% | 19% |
| Jobs over 10 years (destroy more / about equal / create more) | 61% | 27% | 12% |
The report finds optimism about productivity and growth "particularly among wealthier and more educated adults" and pessimism about businesses and jobs "particularly among workers." Asked how many workers replaced by AI in the next decade will be able to learn new job skills in a different industry, 3% said all, 20% most, 51% some, 24% a few and 2% none.
| Policy | Somewhat support | Strongly support |
|---|---|---|
| Creating or expanding workforce training and education programs that teach how to develop or use AI | 41% | 31% |
| Increasing U.S. government spending on AI research | 40% | 16% |
| Increasing government funding for AI research at colleges and universities to hire top AI researchers | 36% | 19% |
| Allowing companies or universities to use government resources for their own AI research | 36% | 13% |
| Creating tax incentives for companies that develop AI | 23% | 8% |
| Reducing rules on AI development to make it easier and faster for companies | 17% | 6% |
Workforce training, at 72% support, was the most popular policy asked about and was, in the report's words, "equally popular regardless of political identification." Tax incentives (31%) and reducing regulation (23%) drew the least support.
Authors' conclusions
The report characterizes public opinion as "a complex web of hopes, concerns, and conditional support." Its policy considerations are the authors' interpretations: that the higher trust among users "suggests that greater adoption may build public confidence in AI technology over time," with "carefully managed deployment and exposure programs" proposed as a way to narrow the gap; that Americans show more support for defensive and deterrent applications; and that the public wants retraining and education programs so that labor-market adjustment does not undermine AI's perceived benefits. The 2026 instalment reported that trust did not rise as use grew, and concluded that higher use and knowledge "are not associated with more positive sentiment" (American AI Sentiment in 2026 (Gallup and SCSP)).
Regulation findings published separately
The report contains no section on rules or safety testing. Gallup published those results from the same survey in a news release of September 16, 2025 by Benedict Vigers and Justin Lall (Source: news.gallup.com):
- 80% said the government should maintain rules for AI safety and data security even if AI capabilities develop more slowly, 9% favored developing capabilities as quickly as possible even if rules were reduced, and 11% were unsure. Support for maintaining rules was 88% among Democrats and 79% among Republicans and independents.
- 97% agreed that AI safety and security should be subject to rules and regulations. On who should create them, 54% named the U.S. government, 53% companies working together on shared rules, and 16% each company setting its own.
- 72% said independent experts should conduct safety tests and evaluations, against 48% for the government and 37% for each company.
- Among those favoring rules, 30% trusted AI somewhat or fully, against 56% among those favoring speed.
- Democrats were nearly twice as likely as Republicans to favor developing AI with a larger group of allies (58% against 30%); 19% of Republicans favored working independently.
Relationships
- instance-of: Public Opinion on AI — the first wave of an annual U.S. opinion series.
- related: US-China AI Competition: Different Races, Different Metrics, AI Race Dynamics — public views on U.S. standing in AI competition, asked with reference to "other countries" rather than China.
- related: American AI Sentiment in 2026 (Gallup and SCSP) — the 2026 instalment that re-measures use, trust and economic expectations.
- related: AI Labor Disruption — 61% expected AI to destroy more jobs than it creates.
- related: Autonomous Weapons, AI and National Security — conditional support for AI-enabled autonomous weapons; expectations of AI-enabled attack.
- related: Gallup, Special Competitive Studies Project (SCSP) — co-producers.