Poll Shows Widespread Public Concern Over AI Development Path, Calls for Stronger Safety Oversight

Multiple recent opinion polls show that public trust in the current direction of artificial intelligence development is declining. According to a nationwide poll conducted in the United States last week by Quinnipiac University and cited by a Twitter user, the vast majority of respondents expressed deep concern about AI's future and strongly support imposing stricter safety constraints on AI companies.

Key Data on Public Concerns

The poll revealed several key figures that paint a picture of the public's true attitude toward AI development:

• 73% of respondents worry that future AI systems may threaten human survival
• 86% support requiring AI companies to meet independent safety standards, even if it slows down development
• 81% believe that in the AI field, safety is more important than staying at the innovation frontier
• 77% advocate slowing or halting development until the safety of powerful AI systems has been evaluated

The overwhelming leanings of these numbers indicate that public concern over AI safety issues has crossed traditional partisan or ideological divides. One detail worth noting: the 86% support rate means that even among groups typically skeptical of regulation, requiring AI companies to comply with independent safety standards has gained broad acceptance.

Quinnipiac University's Polling Center is a well-known independent opinion research institution in the United States, widely cited in American political and public policy research for its relatively neutral survey methodology and high transparency. The institution regularly conducts nationwide sampling surveys covering elections, public health, technology, and other topics, and its data is often used by mainstream media and academic research as an important reference for public attitudes. It should be noted that the results of a single poll can be influenced by questionnaire wording, sample composition, and the timing of the survey. Therefore, when interpreting such results, it is typically necessary to cross-reference them with similar surveys from other institutions to more reliably assess the overall trend in public attitudes.

The Priority Battle Between Safety and Innovation

The most policy-relevant finding in the poll is the clear choice the public made between "safety" and "innovation." 81% of respondents believe safety takes priority over staying at the innovation frontier — a proportion far higher than support for accelerating development.

This attitude stands in sharp contrast to the current mainstream industry narrative. Many AI companies, citing "maintaining competitiveness," advocate for light-touch regulation or even self-regulation. However, the data shows that the public does not endorse this logic of placing speed above safety. 77% support pausing the development of powerful AI systems until safety evaluations are completed — a position even stricter than certain provisions in the EU AI Act.

The EU AI Act is the world's first comprehensive regulatory legislation targeting artificial intelligence, formally adopted in 2024. The act employs a risk-based tiered regulatory framework, classifying AI applications into four categories: unacceptable risk, high risk, limited risk, and minimal risk. For high-risk systems, it stipulates obligations including transparency, data governance, human oversight, and compliance assessment. It is worth noting that the act primarily regulates specific uses and deployment scenarios of AI, rather than universally requiring a pause on the development of powerful AI systems. Therefore, the position supported by 77% of respondents in the poll — "pausing development until safety evaluations are completed" — does indeed exceed the scope of the EU's current legislation in terms of regulatory stringency, coming closer to a cautious approach involving pre-approval or development moratoriums for frontier AI models.

The Absence of Democratic Representation

The poster pointed out a deeper structural problem: although these decisions will profoundly reshape society and the human experience, only a handful of companies and governments are currently steering the direction, with a lack of meaningful democratic participation.

This criticism touches on the core contradiction in AI governance. The power over technological decisions is highly concentrated in the hands of a few tech companies and administrative bodies, while the public affected by these decisions has virtually no institutionalized channel for voicing their opinions. The poll data precisely demonstrates that the public is not indifferent — rather, existing decision-making mechanisms have failed to effectively incorporate their views.

The "democratic deficit" in AI governance refers to the concentration of technological decision-making power in a few tech companies and government departments, while the affected public lacks institutionalized channels for participation. This concept draws on discussions in EU research regarding the legitimacy of technocratic decision-making: when policies involve value trade-offs and interest distribution, expert judgment alone cannot adequately represent diverse public preferences. In the AI field, this deficit manifests as governance frameworks being dominated by industry self-regulation, administrative guidance, and international standards organizations, with public participation often limited to the consultation stage and lacking binding decision-making power. The broad public concern reflected in the poll data precisely demonstrates that incorporating public opinion into regulatory design is not merely a normative issue — it also determines whether policies can sustain ongoing social support.

A Signal to Policymakers

For policymakers, this set of data sends a clear signal. The 86% support for independent safety standards means that mandatory regulation has a solid foundation in public opinion. The 77% support for pausing development until safety can be evaluated provides a mandate for a more stringent, precautionary regulatory path.

If lawmakers continue to ignore this public sentiment, the legitimacy gap in AI governance will further widen. Conversely, incorporating public preferences into the regulatory framework can not only enhance the democratic legitimacy of policies but may also provide companies with clearer, more stable compliance expectations.

Conclusion

The trajectory of AI development should not be determined by a handful of companies and nations alone. The Quinnipiac poll shows that the public has already formed a clear position: safety first, independent oversight, cautious advancement. Excluding these demands from the decision-making process not only violates democratic principles but may also deepen social distrust of AI technology. How to find a balance between innovation and safety that genuinely represents the public interest will be an unavoidable challenge for AI governance going forward.