LOOKING BACK | How AI Intertwines With Our Politics

By the time the broadcast was over, American politics had been changed for ever.  

John F. Kennedy appeared young, energetic and informed in his television makeup during the debate on that September night in 1960, While an un-made-up Richard Nixon appeared haggard, unshaven and sweaty under the lights. 

By the 1980s and 1990s, school children and political science students were being taught that that first debate by itself had sunk Nixon's campaign and elevated Kennedy to the presidency. The 1960 presidential election is now considered the first election of the television era. 

It may be that the 2026 U.S. midterms, or the presidential election to come, will be thought of as the true first election of the artificial intelligence era. Some are already deeming them so. Then again, some commenters also said the same about the 2024 presidential election, while our eyes are more on the 2028 cycle.  

Whenever the first real AI election is deemed to have happened, it's clear that AI is already significantly impacting our politics. That's not necessarily a good thing or a bad thing, for either mainstream political movement in the United States, or the political right and left globally. 

AI is simultaneously an economic development strategy, a campaign issue, an instrument of political persuasion, a source of misinformation and a subject of growing public distrust. The technology that promised to make information more accessible and decision-making more efficient is increasingly entangled in disagreements over who controls information, whose interests technology serves and whether machines can be trusted to participate in democratic processes. 

The political consequences extend far beyond AI-generated campaign advertisements and fabricated images of candidates. The construction of enormous data centers is provoking local opposition. Concerns about employment, electricity prices and corporate concentration are changing attitudes toward the technology. Researchers are finding evidence of political preferences embedded in popular chatbots, while campaigns are experimenting with AI tools to reach voters and influence their decisions. Meanwhile, technology companies are becoming more politically active as they seek favorable regulations and government support. 

For financial services professionals, these developments present risks that are economic, regulatory, operational and reputational. Political disputes over AI could affect technology valuations, infrastructure investments, energy markets and the regulatory environment governing financial institutions. The technology could also reshape how investors obtain financial information and how clients interpret political developments that influence markets. 

As the country approaches Election Day, a central question emerges: Has AI become another battlefield in America's partisan divisions, or is it creating entirely new political alliances and conflicts? 

AI's Emerging Partisan Divide 

Until recently, support for artificial intelligence did not fit neatly into the traditional American political spectrum. Democrats and Republicans could both embrace technological innovation, economic competitiveness and national security applications while expressing concerns about privacy, employment and government surveillance. That consensus, however, is showing signs of deterioration. 

An October 2026 ABC News/Washington Post/Ipsos poll found that 57% of Republicans considered AI personally beneficial, while 55% of Democrats viewed it as personally harmful. Independents were almost evenly divided, with 50% expressing positive views and 49% negative ones. These findings suggest that attitudes toward AI are increasingly influenced by political identity, although they do not establish that partisanship itself causes those attitudes. 

Research from the Pew Research Center similarly found that Democrats had become more concerned than Republicans about AI's effects on daily life and employment. According to its September findings, 56% of Democrats expressed concern about AI's impact on everyday life, compared with 49% of Republicans. On employment, the respective figures were 75% and 68%. 

These differences are significant, but perhaps more revealing is the extent of agreement across party lines. An AP-NORC survey found that 64% of American adults believed AI was developing too quickly, including 76% of Democrats and 54% of Republicans. Both groups therefore expressed considerable apprehension, even as Democrats were substantially more concerned. 

The emerging division may reflect differences in how Americans interpret the technology's economic and social implications. Republicans are more likely to associate AI development with innovation, deregulation, American competitiveness and technological leadership. Democrats are more likely to emphasize employment displacement, consumer protections, corporate accountability and the potential for technology to reinforce economic inequality. 

These are tendencies rather than rigid ideological positions. Conservative concerns about censorship, surveillance and concentrated corporate power overlap with progressive concerns about discrimination, privacy and monopolistic behavior. Likewise, politicians in both parties support domestic semiconductor manufacturing, national security research and American leadership in advanced computing. 

The political significance of these competing priorities becomes particularly visible when AI moves from software applications into physical communities. 

Data Centers: An Unexpected Bipartisan Backlash 

The rapid expansion of AI infrastructure is generating opposition that often transcends traditional political affiliations. 

Data centers require substantial electricity, water, land and supporting infrastructure. Although developers promise construction employment, local tax revenue and economic development, residents increasingly question whether the benefits justify the environmental consequences, utility costs and disruption to surrounding communities. 

Opposition is emerging in rural and suburban communities, including places where Republican voters might otherwise support deregulation and industrial development. Environmental organizations, neighborhood associations, agricultural interests and electricity customers have found common ground in questioning proposed facilities. 

The political dynamics are especially complicated because AI infrastructure creates economic winners and losers. Technology companies and utilities may benefit from construction and increased demand, while neighboring residents could experience higher electricity costs, water consumption or changes in land use. Local governments must balance anticipated tax revenue against public opposition and infrastructure requirements. 

Research and reporting from UCLA, NPR, Grist and other organizations have highlighted the increasingly bipartisan character of data center resistance. The issue is becoming relevant to state and local elections as candidates confront competing demands from developers, businesses and constituents. 

For investors, this is more than a political curiosity. Data center opposition can affect construction schedules, permitting decisions, power procurement, infrastructure financing and the projected profitability of AI-related investments. 

The same conflicts are emerging over AI regulation. 

Republican leadership generally favors reducing regulatory barriers to innovation, although some conservatives advocate restrictions addressing political censorship, privacy or national security. Democrats have generally supported stronger consumer safeguards and corporate accountability, although Democratic officials in technology-oriented states also seek to encourage investment. 

One major disagreement concerns federal preemption: whether Washington should restrict states from enacting their own AI regulations. Technology companies often favor consistent nationwide standards rather than a complicated patchwork of state requirements. Opponents argue that federal preemption could prevent states from protecting residents against emerging harms. 

Yet support for regulation is not exclusively Democratic. Public Consultation research has found substantial bipartisan support for government oversight of AI. Proposals to slow or restrict advanced AI development have also attracted supporters across ideological boundaries, including figures concerned about catastrophic risks, labor displacement and the concentration of technological power. 

AI therefore presents an unusual political landscape. Americans may disagree about whether the technology is beneficial while agreeing that its development requires oversight. They may support domestic AI investment while opposing construction of a data center in their own communities. 

This complexity makes AI difficult to classify as either a Republican or Democratic issue. 

Is Artificial Intelligence Politically Biased? 

If AI's political reception is becoming partisan, its outputs present another complicated question: Are the models themselves politically biased? 

A growing body of research suggests that AI systems can exhibit political preferences, although the direction, magnitude and significance of those preferences vary substantially depending on the model, the questions asked and the methods researchers use to measure bias. 

A 2025 Stanford Graduate School of Business study examined responses to political questions from 24 large language models developed by eight companies. More than 10,000 American respondents evaluated the models' answers. For 18 of 30 political questions, respondents perceived nearly all the models as leaning left. 

Interestingly, both Republicans and Democrats detected leftward tendencies, although Republican respondents generally perceived stronger bias. OpenAI models received the highest aggregate ratings for perceived leftward bias, while Google's and DeepSeek's models were rated closest to neutral. Even xAI's Grok, developed by Elon Musk's company and marketed as an alternative to politically constrained chatbots, was perceived as relatively left-leaning in this particular study.  

The findings challenge simplistic assumptions about which companies produce liberal or conservative AI. A model developed by an executive associated with conservative politics does not necessarily produce conservative responses, and models from companies perceived as progressive do not always express progressive positions. 

A separate June 2026 Washington Post investigation also found political asymmetries in responses from leading chatbots. Its testing suggested that ChatGPT frequently supplied left-leaning arguments when answering selected political questions. These results are important, but they should be interpreted within the limitations of the study's particular prompts and scoring methodology.  

Other research reaches different conclusions. 

Anthropic published an evaluation in November 2025 comparing the political even-handedness of six AI models. According to the company's methodology, Google's Gemini 2.5 Pro scored 97%, Grok 4 scored 96%, Claude Opus 4.1 scored 95%, Claude Sonnet 4.5 scored 94%, GPT-5 scored 89% and Meta's Llama 4 scored 66%. 

These figures measured how fairly models engaged with opposing viewpoints, not whether they supported Democrats or Republicans. Anthropic's commercial interest in demonstrating Claude's impartiality also makes independent replication important. Nevertheless, the results illustrate that different approaches to measuring political bias can produce substantially different model rankings.  

OpenAI has conducted its own research into political bias, examining approximately 500 prompts across 100 topics. Its October 2025 analysis found that models generally remained close to objective when responding to neutral questions but were more susceptible to political bias when presented with emotionally charged prompts. 

The company identified several recurring problems, including expressing political opinions without attribution, providing unequal coverage of competing perspectives and amplifying a user's partisan language. OpenAI also reported improvements in newer models, although its findings reflect its own evaluation criteria.  

The competing findings reveal a fundamental problem: Political neutrality is extraordinarily difficult to define, much less measure. 

Where Does AI Bias Come From? 

AI systems do not ordinarily acquire political opinions through personal experience or deliberate ideological commitments. Instead, political tendencies can emerge from the information used to train them and the decisions developers make about their behavior. 

Training datasets contain enormous quantities of books, websites, journalism, academic research, social media discussions and other human-produced material. Those sources reflect existing political beliefs, institutional perspectives, historical inequalities and cultural assumptions. 

If certain viewpoints are more prevalent in training material, a model may reproduce those perspectives disproportionately. Yet simply balancing the quantity of conservative and liberal source material would not necessarily produce reliable or impartial answers. Some questions concern verifiable facts rather than equally plausible political interpretations. 

Developers also shape model behavior through supervised training, reinforcement learning, safety policies and system instructions. These processes influence which claims a chatbot emphasizes, which requests it refuses and how it responds to controversial subjects. 

Even seemingly neutral design choices can have political implications. A model that emphasizes environmental risks may appear progressive, while one that prioritizes economic growth and regulatory costs may appear conservative. A chatbot that challenges unsupported factual claims could be perceived as partisan when those claims are closely associated with a political movement. 

Stanford's Institute for Human-Centered Artificial Intelligence addressed this problem in a September 2025 policy brief, arguing that perfect political neutrality is impossible but that developers can approximate neutrality through transparency, balanced treatment and evaluation methods that respect competing viewpoints.  

Bias also depends on how a question is phrased. A chatbot asked to identify the strongest arguments for higher corporate taxes may produce a progressive-sounding response, while the same model asked to explain the disadvantages of taxation may produce a conservative-sounding one. 

Neither response necessarily proves that the system favors a political party. More revealing tests examine whether a model applies comparable standards of evidence, accuracy and scrutiny to competing positions. 

This distinction becomes especially important when users ask chatbots to recommend candidates. 

Recent reporting has highlighted experiments in which Grok and OpenAI models made contrasting partisan selections when forced to choose between candidates. Such demonstrations can reveal tendencies under particular conditions, but they should not be mistaken for evidence that a model consistently endorses an entire political party. 

Political bias is also not necessarily stable. Model updates, changing training procedures, new system instructions and user personalization can alter responses over time. 

Can Chatbots Change Political Opinions? 

The potential consequences extend beyond whether chatbot responses sound liberal or conservative. 

Research indicates that conversational AI can influence users' political attitudes, sometimes without explicitly attempting persuasion. 

Studies from Yale, the University of Washington and other institutions have examined how biased chatbot interactions can shape opinions through selective information, conversational framing and repeated exposure to particular arguments. 

A particularly relevant October 2026 study published in PNAS Nexus found that telling participants an AI model was biased against their political party reduced the model's ability to correct economic policy misconceptions by 28%, compared with a neutral control condition. 

The experiment, involving 2,144 participants, demonstrated that perceptions of political bias can undermine trust even when the AI supplies corrective information.  

This creates a paradox. AI systems may influence political opinions through the information they provide, but accusations of bias may also make voters less receptive to accurate information. 

The implications for financial services are substantial. A chatbot that frames inflation, taxation, government spending, trade policy or financial regulation through a particular ideological perspective could influence how investors interpret economic developments. 

Financial professionals should therefore distinguish between AI-generated factual analysis and AI-generated political interpretation. Models may be useful for comparing policy proposals, but they should not become unquestioned authorities on the economic consequences of elections. 

AI's Growing Role in the 2026 Midterm Elections 

Political bias is only one way artificial intelligence is entering the electoral process. AI is also changing how campaigns produce advertising, communicate with voters, analyze information and conduct political operations. 

Generative AI has dramatically reduced the cost of creating campaign materials. Political organizations can produce written appeals, translated messages, social media graphics, synthetic voices and video content without the production resources traditionally required. 

Campaigns can also use AI to summarize polling, analyze voter concerns, prepare debate materials, monitor public sentiment and generate variations of advertising messages. 

These applications are not inherently deceptive. A campaign that uses AI to edit a speech or translate a policy statement is employing the technology much as businesses use productivity software. 

The risks increase when synthetic content is presented as authentic evidence. 

An October 2026 Axios investigation documented AI-generated attack advertising in several competitive races. Examples included fabricated scenes involving Democratic Sen. Jon Ossoff in Georgia, an artificial family photograph involving Michigan candidate Abdul El-Sayed, and synthetic portrayals of Pennsylvania Gov. Josh Shapiro. 

Other examples involved campaigns or allied groups in New York, Texas and Kentucky. The reporting demonstrated that fabricated political imagery was no longer merely a hypothetical election-security threat.  

Such advertisements raise questions about disclosure and political speech. Should campaigns be permitted to create synthetic depictions of candidates saying things they never said? Should voters be informed whenever AI is used? Does a humorous or satirical advertisement deserve different treatment from a realistic fabricated recording? 

Regulators must balance these concerns against First Amendment protections, the long history of political satire and the practical difficulty of distinguishing ordinary digital editing from AI-generated manipulation. 

AI Is Becoming a Campaign Issue, Not Just a Campaign Tool 

The technology itself is also becoming an electoral target. 

On October 29, Axios reported that the Democratic group Tomorrow Together was preparing a $4 million advertising effort across competitive House districts in Wisconsin, Arizona, Pennsylvania and Michigan. 

The advertisements were designed to criticize Republican incumbents over their support for efforts to restrict state-level AI regulation. The campaign illustrates how a previously specialized debate about regulatory preemption is entering mainstream electoral messaging.  

Meanwhile, opposition to data centers is influencing races in states including Texas, Pennsylvania, Ohio, Michigan, Wisconsin and Arizona. Candidates face questions about electricity prices, water supplies, corporate subsidies and local control of development projects.  

This development may prove more politically consequential than individual deepfake advertisements. AI's impact on utility bills, employment and public infrastructure can directly affect voters who have never used a chatbot. 

Misinformation, Hacking and Election Security 

AI-generated misinformation presents another challenge because sophisticated fabrication tools are becoming accessible to almost anyone. 

Synthetic audio can imitate a candidate's voice. AI-generated video can depict events that never occurred. Automated accounts can produce enormous quantities of persuasive messages, while fabricated local news websites can imitate legitimate journalism. 

These techniques can be used to spread false allegations, distort candidate positions or confuse voters about election procedures. 

The threat is not limited to material encountered directly on social media. Misleading information can enter search results, online databases and the information sources retrieved by AI assistants. 

Investigations involving NewsGuard and other researchers have raised concerns about chatbots citing deceptive or partisan websites when responding to election questions. This creates the possibility of information laundering: unreliable claims gain apparent legitimacy when repeated in polished, authoritative-sounding AI responses. 

Research publicized by Washington State University in October 2026 also drew attention to efforts to target AI information systems themselves. Rather than simply persuading human readers, malicious actors can attempt to influence the material that AI systems retrieve and summarize. 

Cybersecurity threats represent a related but distinct problem. AI can help attackers draft convincing phishing messages, imitate officials, automate reconnaissance and identify software vulnerabilities. Election administrators, political campaigns, vendors and public officials may all be targets. 

However, AI-assisted cyberattacks should not automatically be confused with successful compromises of voting systems. An intrusion into a campaign's email account, the scraping of public election information and unauthorized modification of actual vote records are fundamentally different incidents. 

Maintaining those distinctions is essential to public confidence. 

There is also a danger that AI will be used to undermine trust without changing any votes. Fabricated evidence of ballot tampering, false claims about election equipment or synthetic recordings of officials could circulate rapidly during the period between voting and certification. 

Even when authorities successfully disprove such material, the resulting confusion may persist. 

How Many Voters Actually Use AI for Elections? 

Despite the attention surrounding AI's political influence, its direct use by voters remains relatively limited. 

An October 23 Pew Research Center report found that only 8% of American adults said they were using AI chatbots to obtain information about the 2026 elections. 

Among those users, 76% sought information about candidates' positions, 71% used chatbots to investigate the accuracy of political claims, and 65% researched candidates' backgrounds. 

Notably, 40% sought suggestions about how to vote, while another 40% asked about election logistics, including registration and polling locations.  

These findings provide an important reality check. Chatbots are not yet the primary source of election information for most Americans. 

Nevertheless, their influence could exceed their direct audience. AI-generated material may be distributed through social networks, campaign communications, search engines and traditional media, reaching people who never intentionally interact with AI. 

The distinction between direct and indirect exposure will become increasingly important as researchers attempt to measure the technology's electoral impact. 

Beyond Elections: AI's Impact on Democracy and Political Power 

AI's political influence is not confined to campaigns, voting or the ideological tendencies of chatbots. The technology is changing the relationship between citizens, governments and private institutions that increasingly control access to information. 

One important development is the concentration of technological power. A relatively small group of companies, including OpenAI, Google, Microsoft, Anthropic, Meta and xAI, develops or distributes many of the systems through which consumers obtain AI-generated information. 

These companies influence public discourse through decisions about training data, content moderation, safety restrictions, search rankings and information retrieval. Even when those decisions are made for legitimate technical or commercial reasons, they can have political consequences. 

This concentration creates an accountability problem. Traditional news organizations have identifiable editors, editorial standards and established mechanisms for corrections. AI systems can generate countless individualized explanations without an equivalent public record of how each answer was constructed. 

A second concern is political personalization. AI can potentially allow political organizations to tailor messages to individual voters based on their interests, concerns and behavioral characteristics. Traditional campaigns have used voter segmentation for decades, but AI could make personalization cheaper, faster and more conversational. 

The result could be a political environment in which different citizens encounter substantially different representations of the same candidate or policy. 

A third issue involves the political influence of AI companies themselves. 

As AI investment grows, technology companies have increasingly strong incentives to influence decisions involving regulation, government procurement, energy policy, intellectual property, competition and national security. Their executives, investors and political action committees can become important participants in elections. 

This is not fundamentally different from political activity by financial institutions, pharmaceutical companies or energy producers. What distinguishes AI is that its developers may also operate the information systems citizens use to understand the political process. 

The combination of financial influence and informational influence deserves particular scrutiny. 

The Challenge of Regulating Political AI 

Governments are already attempting to establish safeguards. 

According to the National Conference of State Legislatures, 31 states had enacted laws regulating deepfakes in political messaging as of October 13, 2026. Most rely on disclosure requirements, while others restrict certain deceptive synthetic content during election periods. 

However, the legal landscape remains unsettled. Courts have questioned some restrictions on constitutional grounds, and states differ in how they define deceptive media, parody, prohibited conduct and acceptable disclosures.  

There are legitimate arguments on both sides. Supporters of restrictions contend that voters should not be deceived by fabricated recordings or impersonations. Critics warn that broad restrictions could suppress satire, legitimate political commentary and protected speech. 

An effective regulatory approach must distinguish harmful deception from ordinary creative expression while remaining technologically adaptable. 

Importantly, AI can also strengthen democratic institutions. Election administrators can use it to translate materials, improve accessibility, answer routine questions and support administrative tasks. The National Conference of State Legislatures reports that election officials have begun adopting AI cautiously for relatively low-risk functions, including drafting communications and assisting with ballot preparation.  

The technology can also help journalists investigate public records, researchers identify coordinated misinformation and citizens compare competing policy proposals. 

AI is therefore neither inherently democratic nor inherently authoritarian. Its effects depend on how it is designed, governed and deployed. 

What Financial Services Professionals Need to Know 

For wealth managers, asset managers, financial executives and technology leaders, the politicization of AI creates several overlapping considerations. 

The most immediate is investment risk. Political opposition to data centers, disputes over energy pricing and uncertainty about federal and state regulations can influence the economics of AI infrastructure. Investors with exposure to semiconductor manufacturers, utilities, data center operators, cloud providers and AI-related investment funds should recognize that political risk is becoming an increasingly important component of valuation. 

Regulatory uncertainty is another consideration. Financial institutions deploying AI for investment research, customer service, compliance or financial advice must account for changing requirements involving transparency, consumer protection, data governance and automated decision-making. 

Political bias also introduces operational and reputational risks. A financial chatbot that responds differently to equivalent questions about Democratic and Republican economic policies could undermine client confidence, particularly when discussing taxes, retirement programs, trade policy or financial regulation. 

Financial institutions should test AI applications for factual accuracy, consistency, explainability and political even-handedness. The objective should not be to eliminate legitimate differences between policy proposals but to ensure that models assess competing claims using comparable standards. 

Cybersecurity teams must also prepare for more convincing impersonation and social engineering attacks. AI-generated messages purporting to originate from executives, regulators or political officials could create opportunities for fraud and market manipulation. 

For advisors, perhaps the greatest challenge is separating political narratives from investment fundamentals. 

Elections can influence taxation, government spending, regulation and economic expectations. Yet the emotional intensity of political campaigns often encourages investors to overestimate the relationship between electoral outcomes and long-term portfolio performance. 

AI-generated political commentary could amplify these tendencies by supplying confident explanations that reinforce clients' existing beliefs. 

Financial professionals should emphasize diversified investment strategies, scenario analysis and independently verified economic information rather than allowing partisan AI-generated narratives to dictate portfolio decisions. 

Looking Toward the 2028 Presidential Election 

If 2026 represents a significant expansion of AI's political role, the 2028 presidential election could be considerably more consequential. 

By then, generative video and audio may become more realistic and less expensive. AI assistants may be more deeply integrated into search engines, smartphones, social platforms and workplace applications. Campaigns may have access to increasingly sophisticated tools for analyzing voter sentiment and generating personalized communications. 

More advanced autonomous agents could also expand the scale of political operations, coordinating research, content creation, advertising optimization and voter outreach with less direct human supervision. 

These developments would create opportunities for legitimate political participation and substantial risks of manipulation. 

The most difficult challenge may be distinguishing authentic political communication from synthetic content. As fabricated recordings become more convincing, political actors could exploit uncertainty in both directions: circulating false material as genuine while dismissing authentic evidence as AI-generated. 

This phenomenon, sometimes called the liar's dividend, could weaken accountability even without widespread successful deception. 

Yet technological sophistication does not guarantee electoral effectiveness. Voters may become more skeptical of AI-generated material, campaigns may discover diminishing returns from automated persuasion, and new authentication standards may improve transparency. 

Whether AI strengthens or undermines the 2028 election will depend partly on what governments, technology companies, campaigns and voters learn from 2026. 

Is Democracy at Risk? Is 2026 the First AI Election? 

Artificial intelligence presents genuine risks to democratic institutions, but the evidence does not justify concluding that AI has already overwhelmed the American electoral system. 

The technology can manufacture persuasive falsehoods, influence political interpretations and reduce the cost of coordinated manipulation. It can also improve access to information, facilitate civic participation and help identify misinformation. 

The distinction between the capability to influence elections and demonstrated electoral influence remains critical. Research establishes that AI can affect opinions under experimental conditions, but determining whether those effects change real-world election outcomes is substantially more difficult. 

Nor would it be historically accurate to describe 2026 as the first election involving AI. Machine learning, automated advertising and algorithmic targeting influenced political campaigns long before generative AI became widely accessible. The 2024 presidential election already featured synthetic political media and substantial public concern about AI-generated misinformation. 

What makes 2026 different is the convergence of several developments. AI has become a substantive campaign issue, a technology used to produce political messages, a potential source of electoral information and an industry whose financial and regulatory interests are increasingly tied to political outcomes. 

In that sense, 2026 may reasonably be called America's first comprehensive AI election: not the first election to use artificial intelligence, but potentially the first in which AI's economic consequences, political influence and governance have become interconnected electoral questions. 

For the financial services industry, that distinction matters. AI is no longer simply a promising technology or an investment theme. It is becoming part of the political environment that determines how markets operate, how companies are regulated and how citizens understand economic reality. 

The ultimate question is not whether AI will participate in democracy. It already does. The question is whether democratic institutions can retain meaningful oversight of a technology that increasingly shapes the information, incentives and economic conditions on which political decisions depend. 

And the answer will matter long after the votes of November 2026 have been counted.   

Researched by DWN Staff