Citizens broadly support artificial intelligence regulation and tend to favor safety, public oversight and international coordination over innovation-focused, industry-led or exclusively national approaches, according to research highlighted by The Cryptonomist.
The findings come from a conjoint survey experiment conducted by researchers Magnus Lundgren and Jonas Tallberg across seven countries with differing political and economic characteristics. In a conjoint experiment, participants evaluate policy packages made up of varying features, allowing researchers to estimate which elements influence public support. The cross-national design was intended to determine whether preferences extend beyond a single country or political environment.
The the study found not only general backing for AI oversight but also a recurring pattern in how citizens want that oversight designed. Respondents generally prioritized safety when it was weighed against promoting innovation. They also preferred governance by public institutions to private-sector self-regulation and favored international rules over approaches confined to individual countries.
That combination presents a challenge for policymakers attempting to balance economic development with protection against potential harm. Governments frequently frame AI policy as a trade-off: rules may reduce certain risks, but overly rigid requirements could slow research, deployment or investment. The survey indicates that members of the public may place greater weight on safeguards than much of the policy debate assumes.
The preference for public governance is also significant because companies have played a substantial role in developing AI standards and risk-management practices. Voluntary commitments, internal review procedures and corporate ethics principles can be introduced faster than legislation, particularly while technology is changing rapidly. However, self-governance also raises questions about accountability when the organizations assessing risks are the same ones developing or deploying the systems.
Public regulation can introduce enforceable obligations, avenues for redress and oversight that is institutionally separate from technology providers. It can also subject decisions to legislative and administrative processes. At the same time, effective public supervision requires technical expertise, adequate resources and rules that can be applied as AI systems and uses evolve.
Participants in the study also leaned toward international rather than national AI governance. AI models, digital services and data flows routinely operate across borders, while national laws can differ in scope, terminology and enforcement. International coordination could make expectations more consistent, although agreements among governments can be difficult to negotiate and implement. Countries vary in their legal systems, economic interests, security priorities and tolerance for technological risk.
The study further found that attitudes were connected to perceptions of AI. According to the report, support for safety-oriented governance was strongest among people who regarded the technology as risky, unpredictable or capable of affecting them personally. That relationship matters as AI is used in more consequential settings, including employment, healthcare, finance and law enforcement. In those fields, automated or AI-assisted decisions can shape access to jobs, services and opportunities, making concerns about transparency, accuracy and recourse especially relevant.
Lundgren and Tallberg characterize the broader result as a systematic mismatch between citizen preferences and dominant approaches to AI governance, The Cryptonomist reported. While many existing policy strategies have emphasized room for innovation, national authority and a role for industry self-regulation, the surveyed public tended to favor the opposite choices across those dimensions.
The research does not by itself determine which regulatory model governments should adopt. Public preferences are one consideration alongside technical feasibility, legal constraints, economic consequences and the effectiveness of particular safeguards. Survey responses also cannot substitute for detailed evaluation of legislation or enforcement systems. They can, however, provide evidence about the democratic legitimacy of different policy directions.
The findings arrive as governments and international institutions continue to consider how rules should address AI risks without freezing technology or creating requirements that cannot be enforced. Regulatory questions include who is responsible when systems cause harm, what information developers and deployers must disclose, how independent testing should work and whether obligations should vary according to the risks posed by a particular use.
The seven-country experiment suggests that public expectations should be treated as a central part of that debate rather than inferred from industry or government priorities. If citizens consistently prefer stronger safety protections, public accountability and cross-border coordination, policymakers may face growing pressure to explain why regulatory systems depart from those preferences—or to redesign them accordingly.
Sources: AI regulation