Voters across parties demand AI guardrails and government oversight, per NBC/AIPI poll

A new national poll from the AI Policy Institute (AIPI) and NBC News finds that an overwhelming majority of likely voters across both parties want powerful AI systems to undergo mandatory formal safety reviews before public release. The survey of 1,007 likely voters, conducted June 11-12, reveals a clear demand for government oversight that goes beyond current voluntary frameworks. For AI builders, these findings suggest that regulatory expectations are hardening and could reshape how models are deployed and governed.

What happened

The NBC News/AIPI poll asked respondents to choose between options on safety testing, data center governance, and who should set AI safety standards. Two-thirds of respondents said they prefer AI systems with guardrails over banning AI entirely. However, when the choice was limited to unregulated AI or an outright ban, voters strongly preferred banning AI.

On data centers, 47% of respondents supported allowing them if AI systems had safety requirements, while 38% favored banning them outright. Over 60% of both Republican and Democratic respondents said the federal government should set clear safety standards and evaluate companies' adherence, rather than leaving that role to AI companies. Additionally, more than 80% of respondents agreed that AI companies should not build systems smarter than humans unless they can demonstrate control.

The poll also found a notable partisan shift in trust toward government regulators: Republicans are now more trusting of government regulation of AI than Democrats, reversing the pattern from previous years. A separate Pew Research Center survey reported this month that about two-thirds of Americans think AI is advancing too quickly.

Why AI builders should care

Voters are not asking for a ban on AI. They are asking for rules that are clear, enforceable, and applied before models reach users. The poll shows that the appetite for government-led safety reviews and data center oversight is bipartisan and strong, which affects the environment in which AI products are built and released.

For teams building or deploying frontier models, the practical takeaway is that regulatory uncertainty may shift toward a more structured regime. The current White House executive order on AI cybersecurity focuses on voluntary vetting, but voters want mandatory reviews. That gap between public expectations and current policy could close quickly, especially as primary elections become proxy fights over AI regulation. A recent New York congressional race drew over $40 million in spending from AI-focused political groups.

Practical implications

If formal safety standards for data centers emerge, deployment timelines could lengthen. Builders who rely on large-scale compute may need to demonstrate compliance with security and safety requirements before accessing infrastructure. This could affect training runs for new models, inference pipelines, and agent deployments.

For startups and product teams, the signal is to start building governance processes now. Whether or not a mandatory review regime arrives this year, the trend is clear: future model releases may require documented safety testing, red-teaming results, and adherence to government-set standards rather than self-regulation.

Caveats

This poll represents a snapshot of public opinion, not enacted policy. The White House executive order on AI cybersecurity has not yet formalized a testing mechanism for model safety, and any vetting remains voluntary for AI companies. The openAI GPT-5.6 and Anthropic Mythos 5 access decisions were driven by national security concerns, not a general regulatory framework. Builders should watch for formal rulemaking from the administration and Congress rather than assuming poll results directly predict new laws.

FAQs

What kind of AI safety reviews do experts want before release?

The NBC News/AIPI poll found that an overwhelming majority of likely voters support mandatory formal safety reviews for powerful AI systems before they are released to the public, not just the current voluntary opt-in framework. Respondents preferred having AI systems with guardrails over banning them or leaving them unregulated.

Who supports government oversight of AI and why?

Support for government oversight is bipartisan, with over 60% of both Republican and Democratic respondents saying the federal government should set safety standards and evaluate companies' compliance. The poll found that Republicans are now more trusting of government regulation of AI, reversing a prior pattern, while Democrats remain more skeptical of the government's ability to regulate effectively.

What is the AI Policy Institute's role in measuring public opinion on AI regulation?

The AI Policy Institute (AIPI) is a non-partisan research organization based in Washington, D.C., that conducts polls on public attitudes toward AI governance. The NBC News/AIPI poll of 1,007 likely voters was designed to quantify demand for safety reviews and guardrails across party lines. Peter Wildeford, director of policy at the affiliated AI Policy Network, noted that Americans want to do more on AI safety.

What are Mythos 5 and GPT-5.6 and how do they relate to AI regulation?

These are the latest frontier models from Anthropic and OpenAI, respectively. The government cleared Anthropic to give restricted access to Mythos 5 to trusted partners, while OpenAI released GPT-5.6 to a limited subset of partners due to government safety concerns, instead of a wider public launch. These cases illustrate how national security reviews are already shaping model release decisions.