Consumer AGI Poll Reveals Support for Progress… With Conditions Attached

A new survey suggests that Americans are willing to see artificial general intelligence move forward, but they want the race governed by rules that companies cannot write and enforce by themselves.

The message is especially clear on speed: technical progress has public support only when safety testing, outside review, and meaningful disclosure keep pace with new capabilities.

A survey of 100 Americans found that public opinion on the race to AGI is divided over whether major AI companies will put public safety ahead of competitive pressure. Nearly half of respondents expressed at least a fair amount of confidence in the companies, while just over half expressed little or no confidence. The sharper finding came when respondents were asked what should happen if a company appeared close to developing AGI: 87.6 percent chose slower development, independent oversight, or a complete pause rather than unrestricted speed.

The results describe conditional support for progress. Americans appear open to powerful AI systems, but they expect the companies building them to accept outside scrutiny and follow shared safety rules. A company that reaches AGI first may gain technical prestige, yet that achievement alone will not settle questions about whether the system should be released or how its risks should be managed.

The survey found an almost even split on whether leading AI companies will prioritize safety over the race to develop AGI first:

That distribution matters because the skepticism is more concentrated than the headline split suggests. A small group feels strongly reassured, while a larger group has moved all the way to complete distrust.

The gap between expected capability and trusted judgment is especially visible in perceptions of OpenAI. The company was selected by 31.7 percent of respondents as the organization most likely to develop AGI first. Only 11.4 percent trusted OpenAI most to develop it responsibly. Among respondents who expected OpenAI to win the technical race, just 36 percent also placed the highest responsibility trust in the company.

Google showed a closer match between perceived capability and responsible development. Half of respondents viewed Google as the most likely company to develop AGI first, and 54.6 percent trusted Google most to develop it responsibly. Google was the only company to receive majority support on the responsibility question.

The numbers do not establish why respondents made those choices. Corporate reputation, public controversies, product experience, leadership changes, communication about safety, and familiarity with each brand may all shape the result. They do show that capability and legitimacy are separate forms of public judgment.

Editorial takeaway: Technical leadership needs independent safety evidence before deployment can earn broad public support.

OpenAI’s position illustrates the risk for any company that becomes known for speed more than restraint. If a company leads in capability while trailing in trust, it may face stronger regulatory pressure, slower adoption, and greater resistance from institutions that have to approve or rely on its systems.

A survey of 100 US adults highlights a widening gap between perceived AI capability and public confidence in responsible development. OpenAI was viewed as the most likely company to develop AGI first, yet only 11.4% trusted it most to do so responsibly. Respondents generally supported continued research, but preferred slower progress, independent safety oversight, incident disclosure, shared standards, and the authority to pause risky releases. The findings are directional rather than nationally representative, but they point to a clear challenge: technical leadership will not produce durable public support without visible and enforceable accountability.

Respondents were given four choices for a scenario in which one company appeared close to developing AGI. The results were decisive:

The combined 87.6 percent who selected a constraint represents the clearest signal in the survey. The public’s preferred intervention varies, but the demand for some intervention is widespread.

The figures also show that support for a pause should not be treated as support for ending AGI research altogether. When the responses are grouped by whether development should continue, 48.6 percent favor continued development either quickly or with oversight. Another 51.4 percent favor slowing or stopping it.

That near-even division changes the policy reading. The public is asking for conditional progress. A research program can continue, but its pace and release decisions should depend on safeguards that can be inspected and enforced.

The survey’s findings challenge a simple assumption that public concern comes from unfamiliarity with advanced technology. Respondents were not asked to choose between AI and no AI. They were asked to choose how a powerful system should be developed when the consequences of failure could extend beyond the company building it. Most chose some form of restraint.

Respondents did not agree on a single institution that should decide when AGI is safe enough to release. Government regulators, independent scientific experts, and the company developing the system each received roughly 29 percent support. An international regulatory organization received 12.7 percent.

The larger pattern is more consistent than the individual choices. About 70.5 percent assigned the greatest responsibility to someone outside the company that developed the system. Only 29.5 percent gave the developer the final say.

That result shows the limits of self-regulation in the public mind. Companies have technical expertise and direct knowledge of their systems, but they also have financial incentives, competitive deadlines, and pressure from investors and executives. Those pressures create a reason for independent review even when a company’s internal safety team is acting in good faith.

A workable oversight model could divide responsibilities rather than hand all authority to one institution:

This kind of layered structure would address the survey’s lack of consensus about a single final decision maker. Regulators, scientists, and developers each have information the others need. The important condition is that the developer should not be the only party able to approve its own system.

The relatively low support for an international organization should also be read carefully. It may reflect concern about distant bureaucracy or unclear accountability. It does not necessarily mean respondents oppose countries cooperating on common standards. The strong support for shared safety rules suggests that coordination remains attractive when it is tied to specific responsibilities.

The survey’s most important finding is not that the public rejects AGI; it is that people reject the idea that competitive speed should determine when powerful systems reach the world. The sample is too small to serve as a precise measure of national opinion, but its pattern is consistent with broader skepticism toward companies’ ability to govern AI responsibly. Developers should treat trust as an operational requirement, not a communications problem. Independent evaluation, transparent incident reporting, staged deployment, and real authority to delay a launch would provide stronger evidence than assurances issued by the companies themselves.

Respondents preferred a model in which companies continue competing while following common safety standards. That option received 55.2 percent support, well ahead of the other choices:

When cooperative and coordinated options are combined, they account for 70.8 percent of responses. Purely competitive arrangements account for 29.2 percent.

This preference leaves room for private research, commercial products, and competition over performance. It also establishes a floor that competitors would have to meet. Shared rules could cover testing methods, incident reporting, security practices, model access, red-team evaluations, and conditions for deploying systems with high-impact capabilities.

Common standards would also address a problem that individual company commitments cannot solve. A company that invests heavily in safety can face pressure to cut corners if competitors gain an advantage by taking greater risks. Shared requirements reduce that incentive by making certain safeguards part of the cost of participating in the market.

The standards would have to be specific enough to verify. Broad promises about responsible AI leave too much room for companies to define success themselves. Independent audits, published evaluation results, and penalties for material failures would give the standards practical weight.

The survey found meaningful differences by gender, age, and education. Women expressed lower confidence than men that companies would prioritize safety. Confidence reached 38 percent among women, compared with 61.9 percent among men. Women were also more likely to favor slowing or stopping development, assigning accountability outside the developer, and supporting cooperative structures.

Younger respondents aged 18 to 34 were more likely to expect OpenAI to develop AGI first and more likely to support rapid development. Even in that group, 53.2 percent favored slowing or stopping development. Only a minority of male respondents, 15.2 percent, supported developing AGI as quickly as possible.

Education was associated with different views of company leadership. Respondents with bachelor’s or postgraduate degrees were more likely to expect OpenAI to develop AGI first. Respondents with a high school education or less were more likely to select Google as both the likely leader and the company most trusted to act responsibly.

These differences may reflect work experience, media exposure, familiarity with consumer products, political views, or personal exposure to automation. The survey cannot isolate the cause of each pattern. It can show that a single public message about AI will reach different groups in different ways.

A governance process that relies only on industry experts risks overlooking concerns held by people who will experience AI through schools, workplaces, public services, health care, and financial systems. Broader consultation would improve the quality of safety decisions and make those decisions easier to defend.

The survey included 100 Americans between the ages of 18 and 63. That size allows the results to show patterns and generate questions, but it cannot support the same level of confidence as a large, probability-based national survey.

The available information does not establish that respondents were selected randomly or that the sample matches the national population. As a result, the figures should be treated as an indication of public attitudes rather than a precise estimate of what all Americans believe. A margin of error would be inappropriate without a known probability sample.

That caution does not erase the findings. The major patterns are internally consistent across several questions:

Larger surveys can test whether those patterns hold across the country and among different political, economic, and regional groups. They can also examine how attitudes change when respondents receive more information about AGI, specific safety practices, or the costs of delaying development.

The Pew Research Center has found related skepticism in larger studies. In 2025, 59 percent of the U.S. public said they had not too much or no confidence in U.S. companies to develop and use AI responsibly. Stanford’s AI Index also reported a mixed global mood: the share of people who saw more benefits than drawbacks from AI products rose, while the share who said AI made them nervous also increased.

Together, those findings suggest that support for AI and concern about AI can exist at the same time. People can use AI products, see useful applications, and still demand stronger safeguards for systems with much broader capabilities.

The survey does not show a public demand to end AI development. It shows a public demand for conditions around the development of the most powerful systems. Continued progress has a stronger chance of earning acceptance when companies share safety standards, disclose serious failures, submit to independent review, and accept that release decisions affect people beyond their own organizations.

The winner-take-all model has a basic legitimacy problem. A company can claim that speed is necessary because competitors are moving quickly, while the public sees the same speed as a reason for stronger safeguards. That conflict will grow as systems gain broader abilities and become harder for outsiders to evaluate.

The practical lesson is straightforward: progress needs a safety structure that people can inspect. Companies that build that structure early may face higher costs and slower launches, but they will have a stronger case when asking the public, regulators, and other institutions to accept powerful AI systems.

The survey’s central message is therefore conditional support. Americans are willing to consider the benefits of AGI, while demanding proof that safety decisions are independent, transparent, and enforceable. A technical race can continue under those conditions. Without them, public skepticism will shape the future of AGI long before the technology reaches its final form.