Safety and Control Come First for 42% Choosing an AGI Leader, Consumer Research Reveals

A survey of 100 Americans shows that safety carries more weight than raw capability when people consider who should lead the race toward artificial general intelligence. The results point to a public that is open to progress but unwilling to leave release decisions in the hands of companies acting alone.

The survey captures a sharp trust problem around artificial general intelligence, or AGI. Respondents were asked about which companies might reach AGI first, which companies they trusted to develop it responsibly, and what safeguards should apply if a system appeared close to human-level performance across many tasks.

Their answers were consistent across those questions. Americans want advanced AI research to continue under conditions that include independent oversight, public disclosure of serious failures, and shared safety standards. Only a small minority supported moving as quickly as possible without stronger controls.

The strongest signal came from a direct choice between capability and safety. Forty-two percent selected safety and controllability as the leading priority. Another 44.6% said capability and safety mattered equally. Only 7.9% chose capability alone.

86.6% of respondents placed safety on par with or above capability when evaluating advanced AI development.

That result gives policymakers and AI companies a clear reading of public opinion: technical leadership will carry limited value if the company behind it cannot show that its systems are controllable and that its safety decisions can be checked by people outside the development team.

OpenAI received the highest share of respondents who thought a company would reach AGI first, at 31.7%. The result placed it ahead of several competitors in perceived technical momentum. Yet only 11.4% trusted OpenAI most to develop AGI responsibly.

That gap is important. Public confidence depends on more than a belief that a company can build a powerful system. It also depends on whether people believe the company will slow down, disclose problems, and accept limits when commercial pressure points in the opposite direction.

Google produced a different result. Half of respondents thought Google was most likely to develop AGI first, and 54.6% trusted the company most to develop it responsibly. Google was the only company to receive majority support on the responsibility question.

The contrast suggests that respondents separate technical ability from institutional credibility. OpenAI was viewed as a likely technical leader by many participants, while Google was viewed as a more dependable steward. That distinction could influence how regulators, customers, employees, and business partners respond to future AGI claims.

A company that reaches a major technical milestone without public confidence may face several practical problems:

The survey also found that 15.1% of respondents trusted another company most. This group showed especially strong opposition to the current race. Only 4.8% had meaningful confidence that leading companies would put safety first, and 86.1% wanted development slowed or stopped.

That response does not identify a clear alternative leader. It reflects a broader rejection of the current industry structure, where a small group of large companies controls much of the research, computing capacity, and public conversation about AGI.

A survey of 100 Americans finds that public support for AGI development is conditional on safety, controllability, transparency, and independent oversight. Respondents distinguished sharply between companies they believe may achieve AGI first and companies they trust to develop it responsibly, revealing a leadership gap that technical capability alone cannot close. Although the sample is not a national estimate, its directional signal is clear: most participants reject an unrestricted race and favor enforceable safeguards, outside review, incident disclosure, and shared standards.

Respondents were almost evenly split when asked whether leading AI companies would put public safety ahead of the race to develop AGI. Those expressing a great deal or fair amount of confidence totaled 49.8%. Those expressing not too much confidence or no confidence totaled 50.2%.

The intensity of those answers leaned more heavily toward doubt. Only 15.6% expressed a great deal of confidence, while 20.2% expressed no confidence at all.

This matters because the public’s position changes sharply depending on its level of trust. Among people who had greater confidence in AI companies, 53.1% still wanted stronger independent oversight. Only 16% supported continuing development as quickly as possible.

Among respondents with lower confidence, 71.7% wanted development slowed or stopped. Every respondent who selected stopping development entirely came from this skeptical group.

The practical message is straightforward. Trust would not give companies a license to accelerate without limits. It would create more room for continued research under visible and enforceable safeguards. Without trust, calls for delay and restrictions become much stronger.

This pattern is consistent with broader public opinion research. A 2025 Pew Research Center survey found that 59% of the public had not too much confidence or no confidence in US companies to develop and use AI responsibly. The finding suggests that skepticism toward corporate self-governance extends beyond this smaller AGI-focused survey.

Respondents were given a scenario in which one company appeared close to developing AGI. They then chose how development should proceed.

The options produced a decisive result:

Those figures show that 87.6% favored some constraint on the race. The preferred constraint varied, but the direction was consistent. Respondents wanted a pause, outside review, or a complete halt rather than a race governed solely by competitive pressure.

When the answers are grouped by whether development should continue at all, the split becomes closer. A total of 48.6% favored continuing development either quickly or with oversight. Another 51.4% preferred slowing or stopping.

That distinction matters for policy. Most respondents did not call for an end to AGI research. They called for conditions that would make continued research more acceptable. The public appears to support a controlled path where progress is tied to evidence that a system can be tested, monitored, limited, and shut down when necessary.

The result also challenges the idea that the public will accept risk in exchange for the fastest possible access to more capable systems. Only one person in eight supported that approach in the survey.

The survey points to a practical standard for AGI leadership: companies should be judged not only by what they can build, but by how credibly they can limit, disclose, and correct what they build. Public skepticism is not necessarily opposition to progress; it is a demand that release decisions withstand scrutiny from people outside the commercial race. The sample is small and not demonstrably representative, but its central pattern aligns with broader research showing that many Americans have limited confidence in AI companies’ ability to govern themselves responsibly.

The question of who should decide whether an AGI system is safe to release produced no single winning authority. Government regulators, the developing company, and independent scientific experts each received approximately 29% support. An international regulatory organization received 12.7%.

The larger pattern was clearer than the individual rankings. About 70.5% assigned the greatest responsibility to someone outside the company that built the system. Only 29.5% believed the developer should have the final say.

That result creates a strong case against pure self-regulation. A company may possess the deepest technical knowledge about its own model, but it also faces financial, competitive, and reputational incentives that can affect its judgment. External review provides a way to test claims made by the developer and to examine evidence that the developer may prefer to keep private.

Each outside option carries tradeoffs:

The close split among government, scientists, and developers suggests that many people may favor layered accountability. A company could conduct its own testing, independent experts could review the evidence, and regulators could set minimum requirements for deployment and respond when those requirements are violated.

Respondents were also asked how they would react if a company publicly disclosed a serious safety failure involving an advanced model. Almost half, or 47.9%, said the disclosure would increase their trust. Another 31.1% said it would make no difference. Twenty-one percent said it would reduce their trust.

That gives disclosure a favorable overall result. Seventy-nine percent said transparency would either improve trust or leave it unchanged.

The response is understandable. A serious failure creates concern, but hiding the failure creates a second problem. Public disclosure gives outside researchers, regulators, and affected users a chance to understand what happened and assess whether the company has corrected it.

Transparency works best when it includes useful evidence rather than a carefully worded announcement. A meaningful incident report would explain:

The results also show why disclosure alone cannot repair a damaged relationship. Among higher-confidence respondents, 63.3% said disclosure would increase trust. Among lower-confidence respondents, only 32.6% said the same.

For people who already believe a company acts recklessly, an incident report may be evidence of honesty or evidence that the company allowed the incident to happen. The difference depends on what follows the disclosure. Independent verification, compensation for harm, and enforceable changes will matter more than the announcement itself.

The survey found meaningful differences across gender, age, and education. Women expressed less confidence in AI companies than men. Thirty-eight percent of women had confidence that companies would prioritize safety, compared with 61.9% of men. Women were also more likely to favor slowing or stopping development, with 58.2% choosing that position compared with 44.4% of men.

Women were more likely to assign final responsibility outside the developing company and to prefer cooperative development structures. Even among men, support for unrestricted speed remained limited. Only 15.2% of male respondents favored developing AGI as quickly as possible.

Younger respondents were more likely to identify OpenAI as the probable first developer and more likely to support rapid progress. However, 53.2% of respondents ages 18 to 34 still wanted development slowed or stopped.

Education also shaped perceptions. Respondents with bachelor degrees or postgraduate education were more likely to expect OpenAI to develop AGI first. Respondents with high school education or less were more likely to expect Google to lead and more likely to trust Google to develop AGI responsibly.

These differences may reflect exposure to technology, media habits, professional background, familiarity with corporate brands, or views about regulation. The survey cannot separate those causes. It does show that public opinion is not a single block, and any policy process that involves only industry executives and technical specialists will miss important concerns.

The survey included 100 Americans between the ages of 18 and 63. Its findings are useful for identifying attitudes and testing arguments, but the results should be read with care.

The available description does not establish that the sample was randomly selected or nationally representative. The American Association for Public Opinion Research advises that non-probability samples should not be presented with conventional margins of error. A sample of 100 people can reveal strong patterns among its participants while still differing from the broader US population.

That limitation does not erase the findings. It changes how they should be used. The percentages provide a directional signal about the questions that need better research and public debate:

Larger probability-based surveys could test whether the same patterns hold across regions, income groups, racial and ethnic groups, political affiliations, and levels of technical experience.

These mistakes can turn technical progress into a credibility problem for AGI developers.

The survey’s results suggest that public confidence depends on actions that can be checked from the outside. General promises about responsible innovation will carry little weight if companies retain complete control over testing, incident reporting, and release decisions.

A credible model would give the developer several responsibilities while placing important checks elsewhere:

This structure would create costs. Testing takes time and money. External review can expose weaknesses that affect a product launch. Common standards can reduce a company’s ability to move first. Those costs are part of the price of building systems that may influence employment, education, public services, cybersecurity, and information environments at a large scale.

The public’s position is reasonable under those conditions. People can accept competition when the basic safety rules are clear and enforceable. They are much less likely to accept a race in which each company decides for itself whether its system is safe enough and then asks everyone else to trust the decision.

The survey shows a public that remains open to AGI research while demanding control over how the technology advances. Safety ranked above capability for 42% of respondents and alongside capability for another 44.6%. Nearly nine in ten wanted some constraint on rapid development. More than two-thirds wanted the final safety judgment to involve someone outside the developer.

Those numbers describe a mainstream position with a clear condition: technical progress needs an accountability system that can operate independently of the race itself.

AI companies can respond through public incident reporting, stronger external evaluations, and shared rules that apply to competitors as well as themselves. Policymakers can create review systems that focus on measurable risks instead of broad promises. Researchers can make safety claims easier to inspect and reproduce.

The central issue is whether a company can prove that its safety process remains credible when slowing down would cost it money, market share, or the chance to claim the lead. Until that question has a convincing answer, public support for an unrestricted AGI race will remain small.

The public appears willing to accept ambitious AI development when safety decisions are independent, transparent, and enforceable. That standard gives companies a workable path forward, but it also sets a firm limit on the value of speed alone.