AGI Trust Survey Puts Developer Reputation on the Technology Adoption Agenda

A small survey puts a large question in front of the AI industry: can a company win public adoption when the public doubts its judgment? The answer depends increasingly on developer reputation, independent review, and visible safety practices.

A survey of 100 Americans shows that public trust is becoming a practical condition for advanced AI adoption. Respondents remain open to continued AGI development, yet they strongly reject a winner-take-all race with weak safeguards. Their preferences point toward a specific governance model: companies can compete, while shared safety rules, outside review, and public disclosure set the boundaries.

Artificial general intelligence, or AGI, generally refers to AI systems that can perform a broad range of cognitive tasks at or above human levels. The technology remains a disputed and evolving concept, which makes public confidence especially important. People are being asked to evaluate potential systems before anyone can demonstrate exactly how capable, autonomous, or risky they will become.

The survey exposes a sharp gap between perceived technical leadership and confidence in responsible development. OpenAI was identified as the company most likely to develop AGI first by 31.7% of respondents. Only 11.4% trusted OpenAI most to develop it responsibly.

That 20.3 percentage point gap gives the survey its clearest reputational signal. A company can receive credit for technical progress while losing confidence in its ability to manage the consequences. Among respondents who expected OpenAI to reach AGI first, only 36% also selected OpenAI as the most responsible developer.

Google produced a different result. Half of respondents expected Google to develop AGI first, while 54.6% trusted Google most to develop it responsibly. The two judgments were closely aligned, and Google was the only company that received majority support for responsible development.

The figures suggest that technical leadership and public legitimacy can reinforce each other when audiences see a credible connection between capability and governance. They also show how quickly that connection can break when a company appears to place speed, commercialization, or internal competition ahead of safety.

The survey’s company rankings were distributed as follows:

The 15.1% selecting another company deserve attention. This group showed broad distrust of the current leaders. Only 4.8% of those respondents had meaningful confidence that major AI companies would prioritize safety, and 86.1% wanted development slowed or stopped. The answer appears to represent dissatisfaction with the current field more than loyalty to a specific replacement.

A survey of 100 Americans suggests that public support for advanced AI depends less on technical ambition alone than on the credibility of the companies pursuing it. Respondents remained open to AGI development but strongly preferred slower progress, independent oversight, shared safety standards, and public disclosure over a winner-take-all race. The results also reveal a reputational divide: OpenAI was widely viewed as a likely first developer but received far less confidence for responsible development, while Google’s capability and responsibility scores were closely aligned. Because the sample is small and not established as probability-based, the findings should be treated as a signal that developer conduct, external accountability, and transparent safety practices are becoming part of the adoption decision.

Confidence that leading AI companies will place public safety ahead of the race to develop AGI first was almost evenly divided. A total of 49.8% expressed a great deal or fair amount of confidence. Another 50.2% expressed not too much confidence or no confidence at all.

The intensity of those views leans skeptical. Only 15.6% expressed a great deal of confidence, while 20.2% expressed no confidence whatsoever. This distribution matters because a close overall split can conceal a stronger emotional response among people who distrust the companies involved.

Confidence also changed preferences about what should happen next. Among people with higher confidence in AI companies, 53.1% still wanted stronger independent oversight, and only 16% supported unrestricted speed. Among people with lower confidence, 71.7% wanted development slowed or stopped.

The result creates a narrow path for companies. More trust can preserve support for continued research, yet trust alone does not generate support for an unchecked race. Public confidence appears to function as permission to proceed under conditions, with the conditions carrying much of the weight.

Broader research shows that this concern extends beyond AGI. A 2025 Pew Research Center survey found that 59% of the public had not too much or no confidence in US companies to develop and use AI responsibly. The same view was held by 55% of surveyed AI experts.

The 2026 AI Index public opinion findings show a related tension globally. The share of respondents who said AI products and services offer more benefits than drawbacks reached 59%, while 52% also said AI made them nervous. Support for the technology can grow alongside anxiety about its consequences.

The survey offered respondents a scenario in which one company appeared close to developing AGI. Only 12.4% chose to continue as quickly as possible. The other 87.6% selected a form of restraint:

These figures describe a public preference for controlled progress. When the answers are grouped by whether development should continue, 48.6% favored continuing quickly or under oversight. Another 51.4% favored slowing or stopping.

The distinction matters for policy. A majority preference for slowing or stopping does not mean most respondents reject advanced AI research. Only about one in ten supported a complete halt. The central demand concerns the process of development, especially the prospect of a company making high-stakes decisions while competing to reach a milestone first.

The survey also asked respondents to choose between the most capable system and the safest, most controllable system. Only 7.9% chose capability alone. Another 44.6% said both mattered equally, and 42% chose safety alone. That means 86.6% placed safety on par with or above capability.

For AI companies, this finding has direct product implications. Users, employers, schools, hospitals, and government agencies all need to assess whether a system can be controlled when it fails. A model’s benchmark scores answer only part of that question. Reliability, monitoring, access controls, incident response, and the ability to shut down or restrict a system can determine whether an organization adopts it.

The AGI race is already losing the trust race.

The public is not demanding that advanced AI disappear. It is demanding that no single company gets to decide, under competitive pressure, when a potentially transformative system is safe enough to release. Speed without credible oversight may produce a technical leader that institutions are unwilling to follow.

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

The larger pattern is clearer than the split among the top choices. About 70.5% assigned the greatest responsibility to someone outside the company that built the system. Only 29.5% gave the developer final authority.

That result places a limit on self-regulation. A company can provide technical evidence and operate internal safety programs, yet a public release decision involving potentially general-purpose systems requires review that can challenge the developer’s assumptions and incentives.

A workable system could use several layers of accountability:

This structure also addresses the disagreement over who should have final say. Government agencies, scientists, and companies each contribute different forms of expertise. A layered process can reduce the risks of placing every decision with one institution.

The low support for an international organization does not eliminate the need for cross-border coordination. Advanced models can be trained, deployed, and misused across several jurisdictions. Shared terminology, reporting practices, testing standards, and emergency contacts may prove easier to establish than a single global body with authority over every release.

Nearly half of respondents, 47.9%, said they would trust an AI company more if it publicly disclosed a serious safety failure involving an advanced model. Another 31.1% said disclosure would make no difference, and 21% said it would reduce trust.

The practical message is straightforward: disclosure carries more upside than downside for many audiences. A company that reports a serious failure gives users and regulators information they need to make decisions. It also creates a record that can be checked against later claims.

Transparency has limits. Among respondents who already expressed higher confidence in AI companies, 63.3% said disclosure would increase trust. Among respondents with lower confidence, only 32.6% said the same. Skeptical audiences may view disclosure as a useful first step while still expecting changes in leadership, oversight, testing, or release practices.

A credible incident report should provide more than a general acknowledgment. It should explain:

Companies that share only favorable results leave a credibility gap. Safety claims become more persuasive when outsiders can see how the organization handles evidence that threatens its preferred narrative.

The survey found meaningful differences by gender, age, and education. Women expressed less confidence in companies than men: 38% of women expressed confidence that companies would prioritize safety, compared with 61.9% of men. Women were also more likely to prefer slowing or stopping development, at 58.2% compared with 44.4% among men.

Women assigned responsibility outside the developer at a higher rate as well, 78.4% compared with 62.4% among men. Even among men, support for unrestricted speed was limited. Only 15.2% of male respondents favored developing AGI as quickly as possible.

Younger respondents aged 18 to 34 were more likely to view OpenAI as the probable first developer, at 39.2%, and more likely to support rapid development, at 22.7%. A majority of younger respondents, 53.2%, still wanted development slowed or stopped. Older respondents were more likely to favor outside accountability and coordinated development.

Education was associated with different company perceptions. Respondents with bachelor’s or postgraduate degrees were more likely to expect OpenAI to develop AGI first, at 47.9%, compared with roughly 20% to 28% among groups with less education. Respondents with high school education or less favored Google for both trust and likelihood, with 61.2% trusting Google and 53.8% expecting Google to lead.

These differences may reflect familiarity with technology brands, media habits, professional experience, and political or cultural attitudes. The survey cannot separate those factors. Companies and policymakers should still account for the differences when they explain safety decisions, publish evidence, and design public consultations.

A single trust score can hide groups with sharply different expectations. Adoption research should track who trusts a system, why they trust it, and what evidence could change their view.

The survey included 100 Americans between ages 18 and 63 from across the United States. That sample is useful for identifying questions and patterns, yet it is too small to represent the full US population with precision. The available description does not establish that respondents were selected through a probability-based national sample.

The American Association for Public Opinion Research cautions that non-probability surveys should not report a margin of error. A margin of error describes sampling uncertainty under specific probability-sampling conditions. It cannot correct for an unknown recruitment process or guarantee demographic representation.

This limitation does not erase the survey’s value. The results align with larger research showing public concern about responsible AI development, while offering more detailed questions about AGI competition, external oversight, and transparency. The findings should guide further research and policy discussion instead of serving as a precise measurement of every American’s view.

A larger, probability-based study could test whether the same patterns hold across regions, income levels, political affiliations, occupations, and levels of exposure to AI tools. It could also distinguish between views of consumer AI, workplace systems, military applications, and hypothetical AGI.

These mistakes can deepen distrust even when a company has strong technical capabilities.

The survey gives AI companies a practical agenda. Public trust will respond to observable decisions more strongly than to broad assurances. Several steps can turn that principle into operating practice.

A safety case should explain what the system can do, which risks were tested, which risks remain uncertain, and what controls will apply after launch. It should identify the evidence behind the release decision and the people or bodies who reviewed it.

External review has value when reviewers can access enough information to find problems and when company leaders commit to responding to their findings. Reviewers who can only inspect a carefully limited demonstration will have difficulty earning public confidence.

Companies should establish incident categories, reporting deadlines, affected-party notifications, and follow-up procedures before a crisis occurs. A shared format would make it easier for users, regulators, and researchers to compare performance across developers.

A product team facing a launch deadline should not be the only group deciding whether safety evidence is sufficient. Independent internal safety offices, boards with technical expertise, and outside review can create a meaningful separation between revenue targets and release decisions.

Companies can continue competing on model architecture, efficiency, features, and customer experience while cooperating on baseline safety practices. Shared standards reduce the chance that one firm’s shortcuts become an industry-wide pressure to cut safeguards.

The survey’s central message is reasonable: technical leadership earns limited public support when it arrives without enforceable safety commitments. Americans in this sample continue to see value in advanced AI, yet they expect companies to accept constraints before deploying systems with broad capabilities.

That expectation has consequences for every participant in the AGI race. Developers need evidence that systems are controllable. Regulators need authority to inspect and intervene. Independent experts need access to evaluate claims. Users need clear information about failures and limits. Companies need common rules that prevent competition from rewarding unsafe behavior.

A winner-take-all race offers few of those assurances. Shared standards and independent oversight address the specific concerns raised in the survey while preserving room for research and competition. The arrangement also gives companies a clearer basis for earning adoption: performance can be assessed alongside safety evidence, incident records, and governance quality.

The reputation of an AGI developer will therefore become part of the technology itself. A system’s capabilities may determine what it can do. The developer’s conduct will help determine whether institutions allow it to do those things.