New Study from Program for Public Consultation at the University of Maryland Finds Majorities of Republicans and Democrats Overwhelmingly Favor the Government Regulating Artificial Intelligence

News provided by

Aug 14, 2025, 10:17 ET

Support an International Treaty Banning Lethal Autonomous Weapons

COLLEGE PARK, Md., Aug. 14, 2025 /PRNewswire/ -- The White House recently released its AI Action Plan, which opposes government regulation of AI programs, saying that it would impede the US becoming the dominant force in artificial intelligence globally. A new survey found that bipartisan majorities of Americans, while receptive to the argument that regulation may harm innovation, ultimately favor five proposals under consideration in Congress for the government to regulate AI domestically. An international treaty to ban lethal autonomous weapons was also favored by a bipartisan majority.

"Clearly Americans are seriously concerned about the current and potential harms from AI," comments Steven Kull, director of PPC. "And while the public is wary of government regulation, they are clearly more wary of the unconstrained development and use of AI."

The survey was fielded from July 30th through August 7th, 2025 with 1,202 adults nationwide, by the Program for Public Consultation (PPC) at the University of Maryland's School of Public Policy, and follows an earlier survey PPC conducted in March 2024.

AI That Makes Life-Impacting Decisions
Many AI programs are used to make decisions which can significantly impact people's lives, including in healthcare, banking, hiring, criminal justice, and welfare. Proposals have been put forward to require that such AI programs pass a government test to ensure they do not violate laws, engage in discrimination, or have security vulnerabilities.

The arguments against these regulations were found convincing by majorities of 59-69%, including the arguments that, "being so cautious will hurt innovation," and, "the private sector can move faster than the government to address risks." However, the arguments in favor of regulations did significantly better (77-84%), including that it is better to take a preventative approach than to react after significant harm has already been done.

In the end, bipartisan majorities favored regulating decision-making AI programs, by:

Requiring AI programs to pass a government test before they can be put into use, which would evaluate whether they may violate regulations, make biased decisions, or have security vulnerabilities was favored by 79%, (Republicans 84%, Democrats 81%, very red states 74%, very blue states 85%). Support has barely changed since 2024, when 81% were in support.
Allowing the government to audit the AI programs that are already in use, and require the company to fix any problems that are found, was favored by 78% (Republicans 82%, Democrats 78%, very red states 72%, very blue states 83%) – essentially the same as 2024, when 77% were in support.

One respondent summarized their support for the regulations as follows: "Instead of letting things out into the world and then reacting based on flaws, they must be functioning to their best ability beforehand."

Disclosure of AI Training Data
To aid with the pre-testing and audits, 74% favor requiring companies to disclose information to the government about how their decision-making AI was trained, if requested (Republicans 77%, Democrats 76%, very red states 70%, very blue states 77%). Support has increased only slightly since 2024 (72%).

While the argument that such a regulation is government over-reach that violates companies' rights to privacy was found convincing by a majority, the argument that disclosure is necessary to effectively audit the programs and make sure the data used for training was not collected illegally, did much better.

Deepfakes
AI-generated deepfakes have been increasing in their ability to pass as real, and are increasingly used for political ends. Creating new restrictions on deepfakes registered overwhelming bipartisan support:

Prohibiting the use of deepfakes in political campaign advertisements, such as to depict an opponent saying something they did not, or to depict an event that did not occur, was favored by 80% (Republicans 83%, Democrats 78%, very red states 77%, very blue states 82%).
Requiring that all deepfakes which are shared publicly be clearly labeled as such, was favored by 80% (Republicans 83%, Democrats 81%, very red states 74%, very blue states 85%).

Support for both policies has not moved more than a few percent since 2024, when they were favored by 84% and 83%, respectively.

One respondent explained their support for prohibiting deepfakes in political campaigns as follows: "Using deepfakes for nefarious acts such as swaying elections should not be acceptable. Purposefully putting out misleading videos and pictures to the public as the truth will be the downfall of any democratic society."

Lethal Autonomous Weapons
The UN Secretary General recently called on nations to come together to ban the development of weapons that can use AI to fire on targets without human control, known as lethal autonomous weapons. The US government has so far been opposed to such a ban, and has been developing its own autonomous weapons since 2023 as part of the Replicator program.

The arguments against such a treaty were found convincing by majorities (60-64%), including that lethal autonomous weapons have the potential to reduce civilian casualties, and the US placing constraints on itself and its allies could give an upper hand to enemies that decide to use such weapons. The arguments in favor did better though (74-81%), including that the risk of such weapons getting out of control is too high, and such risks will never truly be known until they are deployed.

In the end, a large majority of Americans (74%) favored the US working to create a treaty banning the development of lethal autonomous weapons, and creating an international agency to enforce that prohibition. Support for the proposal was overwhelmingly bipartisan, with no significant difference between the parties (Republicans 75%, Democrats 77%, very red states 71%, very blue states 76%). Since 2024, support has declined a little, from 81% to 74%.

About the Survey
The survey was a "public consultation survey" in which respondents are provided briefings and arguments for and against proposals. Content was reviewed by experts from each side of the debate to ensure that the briefings are accurate and balanced and that the arguments presented are the strongest ones being made.

The survey was fielded July 30th through August 7th, 2025 with 1,202 adults nationally. Samples were obtained from multiple online opt-in panels, including Cint, Dynata and Prodege. Sample collection and quality control was managed by QuantifyAI under the direction of the Program for Public Consultation. Samples were pre-stratified and weighted by age, race, gender, education, income, geographic region, marital status, and home ownership to match the general adult population. The survey was also weighted by partisan affiliation. The survey was offered in both English and Spanish. The confidence interval is +/- 3.0% and the response rate for the sample was 7.9%.

Classification of states as "very red" and "very blue" are based on the Cook Partisan Voting Index: The quarter of respondents who live in the most red states were categorized as "very red," and vice versa for "very blue."

CONTACT: Allison Stettler, [email protected]

SOURCE Program for Public Consultation