AGI Research Reveals a Confidence Gap as 50% Doubt AI Companies Will Put Safety Before Winning

A survey of 100 Americans shows a public willing to live with continued AGI research under strict conditions. The strongest message is about process: companies can keep competing, but safety decisions need independent review, public accountability, and rules that apply across the industry.

A new survey captures a sharp confidence gap in the race toward artificial general intelligence. About half of respondents expressed confidence that leading AI companies will put public safety ahead of the effort to reach AGI first, while 50.2% expressed little or no confidence. The result leaves the public divided over corporate intentions and far more unified about the safeguards required before powerful systems move forward.

The survey included 100 Americans between ages 18 and 63 from across the United States. Its findings offer a useful snapshot of public attitudes at a time when Google, OpenAI, Meta, Anthropic, and other companies are investing heavily in increasingly capable systems. The responses show conditional support for development, paired with strong resistance to a winner-take-all race governed mainly by the companies building the technology.

That distinction matters. Public skepticism centers on how AGI is developed, tested, and released. Technical progress can earn attention, but public legitimacy depends on evidence that safety decisions are independent, transparent, and enforceable.

Nolla Health began a one-year Utah pilot focused on adults with mild to moderate acne. Patients pay $4.99 for the subscription, submit health information and facial photos, and receive an AI-generated treatment recommendation that can be sent to a pharmacy. Licensed clinicians remain available through telehealth, and cases involving drugs outside the program’s topical treatment list are routed to them.

The arrangement gives Utah a useful testing ground for a question that reaches well beyond dermatology: how much medical decision-making can be automated safely when the system has a limited scope, a low price, regular follow-up, and a clear path to human care?

Respondents were asked how much confidence they had that leading AI companies would prioritize public safety over the race to develop AGI first. The results divided almost exactly down the middle:

The intensity of the answers adds important context. Strong confidence was less common than complete skepticism. That pattern suggests a fragile form of public acceptance. Many people may be willing to see research continue, while expecting companies to prove that their incentives are aligned with public safety.

The split also shaped attitudes toward the next stage of development. Among respondents with higher confidence in companies, 53.1% still wanted stronger independent oversight. Only 16% supported developing AGI as quickly as possible. Among respondents with lower confidence, 71.7% wanted development slowed or stopped.

Confidence therefore changes the preferred level of restriction rather than removing the demand for oversight. People who trust companies more tend to support continued development with controls. People who trust them less tend to support delay or a halt. The common point is the need for a process that can check corporate decisions.

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.

The responses about individual companies reveal an even wider gap between perceived technical leadership and responsible stewardship.

OpenAI was viewed by 31.7% of respondents as the company most likely to develop AGI first. Only 11.4% trusted OpenAI most to develop AGI responsibly. That is a 20.3-percentage-point difference between expected capability and confidence in judgment.

Among respondents who predicted that OpenAI would reach AGI first, just 36% also selected it as the company they trusted most for responsible development. The rest placed their trust elsewhere, with Google receiving 33.1% among this group and other responses spread across Meta, Anthropic, and unnamed alternatives.

The survey captures the resulting attitude in a blunt sentence:

“We think you will get there first, but we do not trust you to do it right.”

That sentence describes a serious legitimacy problem. A company may lead on benchmarks, product releases, or investment while still struggling to convince the public that it can manage high-consequence risks. Technical achievement and institutional credibility develop on different tracks.

Google occupied a stronger position on both measures. Half of respondents viewed Google as the most likely company to develop AGI first, while 54.6% trusted Google most to develop it responsibly. Google was the only company to receive majority trust for responsible development, and its perceived technical lead broadly matched its trust score.

Survey responses cannot establish why these perceptions differ. Corporate reputation, public communication, leadership disputes, safety controversies, product experience, and familiarity with each company may all contribute. The practical lesson is clearer than the underlying cause: companies need to earn confidence through visible governance, rather than expecting technical success to create trust automatically.

The 15.1% who selected another company offer a separate warning. This group showed very limited confidence that leading companies would prioritize safety, with only 4.8% expressing meaningful confidence. Some 86.1% wanted development slowed or stopped. The answer appears to represent broad dissatisfaction with the current group of industry leaders rather than strong support for a specific alternative.

The survey presented a scenario in which one company appeared close to developing AGI. Respondents then selected the approach they preferred:

The combined figure is the central finding: 87.6% preferred some constraint on the AGI race. The most popular option involved slowing development until safeguards were stronger. The second most popular option allowed continued progress under independent oversight.

The results also show that public concern has limits. When the responses are grouped by whether development should continue at all, 48.6% favored continuing either quickly or under oversight, while 51.4% favored slowing or stopping. Only about one in ten supported a complete halt.

This is a mandate for conditional development. The public appears willing to accept research and competition when companies meet defined safety requirements. A race with no independent checks has little public support.

That preference should affect how industry leaders describe competition. Speed can remain one measure of success, but release readiness needs to include evaluation quality, incident reporting, controllability, and the ability of outside reviewers to challenge a company’s decision.

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 were also asked who should have the final say about whether an AGI system is safe to release. Three options received roughly equal support:

An international regulatory organization received 12.7%.

The distribution leaves one broad conclusion. About 70.5% assigned the greatest responsibility to someone outside the company that built the system. Only 29.5% gave the developer final authority.

The responses do not identify a single preferred regulator. They do show a strong preference for shared accountability. Government agencies could set legal requirements. Independent experts could test claims and publish evaluations. Companies could remain responsible for engineering, monitoring, and incident response. An international body could coordinate standards across borders where appropriate.

This layered approach fits the public mood better than pure self-regulation. A company that builds an advanced system has detailed technical knowledge, but it also has commercial pressure, competitive deadlines, and a direct interest in release. An external review process can add a second judgment before deployment decisions become irreversible.

The low support for an international organization deserves careful interpretation. It may reflect concerns about distant institutions or unclear authority. It does not necessarily indicate opposition to cross-border cooperation. Shared testing methods, incident reporting rules, and minimum safety requirements can operate across countries without placing every release decision in one global institution.

Respondents were asked to choose between building the most capable AI system and building the safest and most controllable one. The answers placed safety at the center of the development goal:

Combined, 86.6% placed safety on par with or above capability.

This result explains why the unrestricted race has so little support. A system that performs impressive tasks while remaining difficult to control does not meet the public’s preferred definition of progress. Capability creates value only when people can understand, limit, and correct the system’s behavior.

For technical teams, this means safety needs measurable release criteria. Those criteria could include resistance to misuse, reliable shutdown behavior, protection against unauthorized access, monitoring for dangerous capabilities, and independent testing under realistic conditions.

The public may never agree on the exact threshold for AGI safety. A workable governance process can still define what evidence companies must provide before release. That evidence should cover both normal operation and failure conditions.

The survey’s preferred competitive structure was a hybrid model. Some 55.2% favored companies competing while cooperating on shared safety standards. Other responses favored close competition among several companies at 20.2%, government coordination at 11.2%, one company taking a clear lead at 8.9%, and international management at 4.4%.

When cooperative and coordinated options are combined, 70.8% favored some form of collective structure. Purely competitive arrangements received 29.2%.

This preference offers a practical path for policymakers and industry leaders. Companies can continue to compete over model design, applications, efficiency, and commercial performance while agreeing on baseline safety requirements. Shared standards can cover areas such as:

The key issue is enforceability. Voluntary standards can provide a starting point, especially when regulators and technical experts need time to develop detailed rules. Companies must face meaningful consequences for ignoring agreed requirements, concealing incidents, or releasing systems that fail independent evaluations.

A shared standard also reduces the competitive pressure to cut safety work. If every major company must meet the same baseline, one firm gains less by treating safety as a cost to be avoided. Competition can then focus more heavily on engineering quality and useful applications.

The survey is small, but its themes overlap with larger public opinion research. A 2025 Pew Research Center survey of 5,023 US adults found that 59% had not too much or no confidence in US companies to develop and use AI responsibly. Pew’s question and sample differ from the AGI survey, so the figures should not be combined. The two studies point toward a similar public concern about corporate responsibility.

Stanford University’s 2026 AI Index report found a more mixed global picture. The share of respondents who said AI products and services offered more benefits than drawbacks rose to 59% in 2025, while 52% said AI made them nervous. Those figures describe an audience that can see practical value and feel substantial unease at the same time.

The pattern is consistent across these sources: people are evaluating AI through both utility and risk. Positive experiences with AI do not automatically translate into confidence in the companies developing more powerful systems.

The survey found substantial differences by gender, age, and education. Women were more skeptical than men about companies prioritizing safety. Confidence reached 38% among women and 61.9% among 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 assigned responsibility outside the developer at a higher rate as well: 78.4%, compared with 62.4% among men. They also showed stronger support for cooperative development structures.

Age differences were smaller. Respondents ages 18 to 34 were more likely to view OpenAI as the probable AGI leader, at 39.2%, and more likely to support rapid development, at 22.7%. A majority of younger respondents, 53.2%, still favored slowing or stopping development.

Education was associated with different views of the major companies. Respondents with bachelor’s or postgraduate degrees were more likely to expect OpenAI to develop AGI first, at 47.9%. Respondents with high school education or less were more likely to select Google for both trust and expected leadership, with 61.2% trusting Google most and 53.8% expecting Google to reach AGI first.

The reasons behind these differences require more research. Familiarity with products, employment in technology, news consumption, political attitudes, and personal experiences with automated systems may all influence responses. For companies and policymakers, the practical lesson is that a single public message will reach different groups differently. Clear evidence and independent review can travel across those divisions better than branding alone.

The findings are useful as a directional signal, with a sample of 100 respondents and limited demographic detail in the published account. A sample of this size can show strong patterns within the group surveyed. It cannot establish precise national opinion by itself.

Sampling method also matters. The American Association for Public Opinion Research advises researchers to avoid reporting a conventional margin of error for a non-probability sample. Any interpretation of these results should therefore focus on the size and consistency of the preferences rather than treating each percentage point as a precise estimate of all Americans.

The strongest findings remain meaningful despite those limits because they are large and internally consistent:

A larger, probability-based survey could test whether those patterns hold across the national population and examine how views change when respondents receive more information about AGI risks, benefits, and governance options.

The public has set a clear condition for continued progress. Companies can compete for technical leadership while accepting outside scrutiny, shared standards, and meaningful limits on release decisions.

That condition serves the industry as well as the public. A company that reaches AGI first but lacks credibility may face delayed deployment, tighter regulation, employee opposition, customer hesitation, and costly crisis management. A company that builds strong independent review into its process can move more slowly in the short term while gaining a stronger foundation for adoption.

The survey’s message is therefore practical. Public trust will follow visible behavior. Companies need to show who can challenge their safety claims, what evidence supports a release, how incidents are disclosed, and what happens when a system fails. Policymakers need to create rules that prevent companies from serving as the sole judge of their own risk.

The AGI race can retain competition and ambition under those conditions. Its credibility depends on whether safety decisions are open to examination and backed by consequences. Technical leadership carries lasting value when the public can see that the system behind it deserves trust.