Figure 3. Percent of U.S. voter parents who reported each worry about their children’s future

Furthermore, a significant share of voting parents with children at home are concerned about how AI will affect their kids’ future. We find that 39% of registered-voter parents nationwide worry about how technology is changing childhood (admittedly, that includes a broader category of technology, not just AI). More specifically, 38% fear what AI means for the future job prospects of their kids. 

At first pass, these percentages might appear to be somewhat low, and, indeed, parents are more worried about childhood health and wellness (51%), safety from crime (50%), or education quality (47%). But still, concern for AI’s effect on children’s future careers is on par with concern for the political trajectory of our country (38%), not far below the priority placed on concerns about children’s mental health and happiness (42%), and is a bigger concern than worries about society’s morals (33%). It is also significantly more important to parents than concerns about the transmission of religious practice (26%), as well as concerns about climate change (21%). 

Moreover, what separates AI from all these other issues is its newness. Most parents only became aware of AI with the release of ChatGPT a mere three years ago, or even more recently. In other words, AI has skyrocketed from not even being on the parental radar, to now being a major concern of parents. And unless the trajectory substantially changes, we expect the relationship between AI and families to grow more fractious over time.
 

Figure 4. Percent of registered-voter parents who reported each worry for their children’s future, by partisanship 

It's not just their children’s careers that Americans worry about. They worry for their own jobs, too. We wanted to know how Americans felt about their job security in the age of AI, especially within a very tight time horizon of the sort envisioned by those who expect artificial general intelligence before 2028. So, we asked respondents how concerned they are that AI will take their job “in the next two years.”

Given that extremely short time span, we assumed that the largest share of our sample would not be worried at all—which is exactly what we found. In fact, 47% of American voters say they are not worried at all that AI will take their job over the next two years. But 27% are fairly or very worried that they will be imminently displaced; and 24% are at least a “little worried” (2% of our sample claims to have already lost their jobs to AI). In other words, many Americans experience AI as a source of economic precarity, and given the time horizon, an intense one at that. 
 

Figure 5. Percent of registered voters who reported each employment worry level

Concerns about AI replacement are hardly random: college-educated voters are a lot less worried about AI taking their job (though a significant share, approximately 1 in 5, do worry).
 

Figure 6. Percent of U.S. voters who were fairly or very worried about AI taking their jobs by 2028, or who reported already having been replaced, by education 

Highly-skilled workers may be likelier to see AI as an extra tool in their toolkit, or what economists describe as a “complement” to their labor, while less-skilled workers—having experienced innumerable corporate techniques to seek cheaper labor—view AI as likely to replace their jobs.

What Do Red States Think About AI Liability?

In September 2025, we conducted a poll that found that Americans are overwhelmingly supportive of regulations to penalize AI companies for harms to kids and consumers. We wondered if our new poll, conducted with a different sample by a different company and using different questions, would find similar results. As it turns out, the answer is yes. 

We did not ask a full roster of AI policy questions but instead tested their popularity against other, more established policies about how to govern technology—such as age verification for pornography sites, and bell-to-bell removal of smartphones from schools. We find that support for the regulation of AI companies is similarly popular to these more tested ideas (and morepopular in several cases).

Figure 7. Percent of U.S. voters likely to vote in 2026 midterms who supported each policy

Nationwide, about 80% of respondents want Congress to hold AI companies legally liable for harms to children. Furthermore, 62% of American voters believe that state governments should be free to regulate the use of AI in businesses and at home; and 76% of voters agree that the federal government should pass laws requiring employers to report when a layoff was caused by the deployment of AI. 

Nationally, we find strong support among both Republican-aligned voters (i.e., those who voted for Trump in 2024, or didn’t vote but are conservative) and Democratic-aligned voters (i.e., those who voted for Harris in 2024, or didn’t vote but are liberal) for policies to regulate AI.  

Figure 8. Percent of U.S. voters who support each policy, by partisanship

Support for regulation of AI is bipartisan, and, likely, a political winner—perhaps even politically unifying. 

We then drilled down into our key states: Florida, Kentucky, Louisiana, Tennessee, and Utah (Red), as well as Michigan (Purple). The results are roughly equivalent to the national numbers above, if not stronger.
 

Figure 9. Percent of U.S. voters show support each policy, by state

But we also looked more closely at what Americans think AI companies should be legally responsible for, so we provided respondents with five scenarios for holding companies liable.
 

Figure 10. Percent of U.S. voters in each state who support imposing “major financial fines or penalties” on companies, by given scenario

Respectively, 70% of American voters think companies should be held liable for convincing a depressed teen to commit suicide, 58% agree that they should be liable for causing a major error in a legal case, 69% support liability for an AI system going rogue and shutting down a powerplant, 70% for giving instructions to a terrorist, and least of all (but not negligible), 43% believe they should be liable for giving a student incorrect study help. 

An analysis of our six key states (as shown in the figure above) shows that voters support holding AI companies liable for these various harms at a few percentage points above or below the national average. In fact, in most instances, the states scored above the national average. These red states tend to be more supportive of holding AI companies liable for harms than the national average, especially in the case of AI convincing a teen to commit suicide, as well as facilitating catastrophic public harms, such as shutting down a power plant or assisting in the plan of a terrorist attack.  

AI and Political Candidacy

We have shown that voters have negative views of AI generally, see AI as a threat to their careers and as a challenge to family life and parenting, and regard accelerationist rhetoric with skepticism. We have also shown that voters generally favor a much more intense regulatory regime around AI, imposing significant liabilities on companies when their products give damaging advice to individuals. But does a politician’s stance on AI actually matter for voter choice? Yes, it does. 

To answer this question, we conducted a randomized controlled survey experiment. We asked respondents if they would be more or less likely to vote for the Democratic or Republican candidate in their district in 2026 if they found that the candidate: supported laws making AI companies liable for harms to children vs. supported laws limiting state AI regulation, vs. supported accelerated permitting for power generation for AI data centers. Each respondent was given just one combination of candidate partisanship and policy stance. Pooling all these responses enabled us to see how a candidate of a given party can expect voters to respond to their stance on a given issue. 

While respondents’ reports about vote-shifting should be taken with a grain of salt, because we randomly assigned candidate partisanship and the specific issue stance, our results clearly show whether a pro-AI or anti-AI stance has more benefits or liabilities for elected officials. In other words, the results show genuine causal estimates of how a change in a given candidate’s stance might relate to how a voter views them, positively or negatively. 

Our analysis begins on the national level and then bores down to the level of states, eventually differentiating between voters in red, blue, and purple states, as well as between Trump and Harris voters. We were ultimately interested in the results of red states and Trump voters, as their opinions respond to policies being generated by Republican leadership. Our hypothesis about Trump voters was that—given existing dynamics of political polarization—they would not be motivated to switch their votes to Democrats in blue and purple states, where any Republican candidate would be preferable to them. But what about Trump voters in red states: would a choice between pro-AI vs. anti-AI candidates yield a switching of votes? As we shall see, the answer is yes: Trump voters in red states strongly support candidates that oppose AI companies, and not vice versa.

Figure 11. Percent of U.S. voters saying stance would make them more likely to support candidate minus percent saying stance would make them less likely, by stance and state

Focusing first on the six key states, voters in every state we surveyed in depth are similar to the overall national average in being supportive of political candidates whose policies hold AI companies liable for harms to kids and oppose preemption, as well as cutting red tape to open AI power plants. Though Kentuckians were perhaps slightly less anti-AI, and Louisianians, Utahns, and Michiganders slightly more so.
 

Figure 12. Percent of U.S. voters saying stance would make them more likely to support candidate minus percent saying stance would make them less likely, by stance, state, and political background

Within states, we find some significant differences between Democratic-aligned and Republican-aligned voters, as well as between Republicans nationally and Republicans locally. As this figure shows, Trump and Harris voters nationally, as well as in the six key states, support candidates that make AI companies liable for child harms. 

On the whole, both Democratic and Republican candidates can expect that supporting stricter liability for AI companies whose products harm children is a winner with voters, and particularly with “base” voters who are most active in midterms. For example, Harris voters strongly favor Democratic candidates who will regulate AI companies, and Trump voters strongly favor Republican candidates who will regulate AI companies. 

But there are differences for preemption and for the addition of new power plants. While Republicans nationally had fairly neutral views of these stances, Republicans in the specific states where we collected larger local samples had more negative views. Especially in Michigan and Louisiana, Republican-aligned registered voters have a very strong willingness to shift their vote against candidates who support federal preemption.

Because of the large difference between Republican-aligned voters in our key focus states and Republican-aligned voters elsewhere, we extended our analysis to cluster our respondents into three groups: the 17 states where President Trump received the highest vote shares in 2024 (red states), the 16 states and DC where he received the lowest vote shares (blue states), and the 17 states where he had intermediate shares (purple states). In each group, we assessed how individual-level partisanship influenced candidate support.

Figure 13. Percent of U.S. voters saying stance would make them more likely to support candidate minus percent saying stance would make them less likely, by stance, candidate party, respondent political background, and state group

The above figure shows the influence on hypothetical candidate support for candidates of a given party, in states of a given political climate, for voters of a given partisan affiliation. Among Democratic-aligned voters and for Democratic candidates, there are not usually huge differences in AI-policy-views across state political contexts. 

But on one issue, there is real variation across contexts: federal preemption. Republican candidates in red states face large political penalties if they support federal preemption. Both Democratic- and Republican-aligned voters in the states President Trump won by large margins report very negative views of pro-preemption Republican candidates. Republican candidates in blue and purple states are not penalized, but red-state Republican candidates may face very serious political costs if they come out in favor of federal preemption. 

We should remind readers that this experiment was designed to analyze how voters would react to different candidates under a variety of scenarios. Overall, our prior research, as well as this survey, shows that the idea of preemption remains unpopular nationally. Whatever the case, the overall preference of Republican and Democratic voters alike is clear, both nationally and in these six states, for candidates that will hold AI companies liable for harming kids, oppose federal preemption, and not give AI companies special regulatory carve-outs. While exact nuances on preemption may vary, there are absolutely no pro-AI positions that are winners for elected officials.

Conclusion

Overall, the big picture painted by our findings is decidedly bad for the AI industry. For the large majority of American voters, AI is a source of concern and fear (and not hope). Americans strongly agree with statements that call for robust protections against AI, and they support leaders who see AI as a negative force over those that call for its unrestrained expansion. Americans are growing increasingly uncomfortable with the concrete presence of AI in their lives, especially parents and workers, for whom it is a growing source of concern and precarity. Finally, American voters support various policies to protect them in the age of AI and want AI companies to be held liable for harms to children and for other catastrophes. 

Most Americans believe that AI companies should be penalized for destructive uses of AI, and they’ll vote for candidates who agree. In practice, Americans think any companies offering “intelligence,” whether human or artificial, necessarily incurs the moral and legal duties that accompany such intelligence. These dynamics portend trouble for any political party that advances AI policies that favor Big Tech companies without offering robust regulatory safeguards as well.

But as for the last question of this brief—will there be electoral repercussions for accelerationist politics—we sought to be extremely careful in our prognostications. The short-term indicators vary across voter ideologies, candidate affiliations, and state political contexts, and across red states, blue states, and purple states, especially as it pertains to the politics of preemption. 

While legislators who seek to protect children have overwhelming bi-partisan support, Democrats strongly oppose candidates that accelerate electricity generation for data centers and support federal preemption. In the aggregate, Republicans, by contrast, are more mixed on these issues. Though there are signs that accelerated permitting for data centers is growing as an electoral issue, our survey finds that this is not yet a national issue for Republicans (but that there may be, at most, some extremely modest negative effects for candidates that support the opening of data centers in red states). This overall picture is probably the result of Republicans simply being supportive of streamlining regulation and development generally, but their feelings may shift more dramatically in the long-term if it ends up effecting their energy bills (as some argue it will). Time will tell if this becomes an electoral issue for Republicans in the years ahead.

On candidates who support federal preemption, the opinions of Republicans are more sharply shaped by local political context. In blue states, they favor Republican candidates that support preemption; but in red states, they oppose them. In other words, Republican-aligned voters in the red states who most reliably send Republicans to Congress, and where right-wing primary threats may be most potent, are strongly opposed to candidates who support federal preemption.

Thus, our results range from cases where the public overwhelmingly opposes the interests and arguments of the AI industry, to cases where it is at best ambivalent towards them. This can be seen as evidence against the viability of AI accelerationism as a salient political force in the United States. Policymakers advancing this view will likely pay electoral costs, perhaps sooner rather than later. With the 2026 midterms approaching, it is unclear which candidates and parties are aware of these costs. Those who ignore them may find themselves unexpectedly thwarted at the ballot box.

Editor's Note: Download the full research brief for footnotes.