The Regulatory Vacuum and Who Benefits From It

The received explanation for the absence of meaningful federal regulation of AI in elections is regulatory lag: technology moves faster than legislatures, democratic institutions are slow by design, and the AI revolution has simply outpaced the rulemaking processes designed for a different era. This explanation is comfortable because it is partly true and because it implies the solution is patience — that regulation will eventually catch up, as it has with previous technological disruptions.

It is also a significant misdiagnosis.

The absence of AI election regulation in the United States in 2026 is not primarily a product of institutional slowness. It is the product of a specific political economy in which the entities most capable of writing effective rules — technology companies and well-funded political campaigns — have strong financial interests in the absence of those rules, and in which the political actors who would pass the rules are among the primary beneficiaries of the unregulated environment.

This is the structural feature that makes AI election regulation genuinely hard, as distinct from merely technically complex or institutionally slow.

The Regulatory Lag Story and Its Limits

The regulatory lag argument has a genuine empirical basis. The FEC’s rulemaking process for AI in political advertising, initiated in August 2023, had not produced final rules by the 2026 election. Three years from initiation to final rule is not unusual for FEC rulemaking — the agency has a bipartisan structure that requires agreement across party lines, and Republicans and Democrats have consistently disagreed about the scope and advisability of AI advertising regulation.

But the regulatory lag story stops making sense when you look at what has happened when the political incentives align.

The Senate passed the DEFIANCE Act — addressing non-consensual deepfake intimate images — in July 2024, from introduction to presidential signature in under three months. That’s fast by congressional standards. The policy area was AI-generated content. The regulatory target was clearly defined, the harm was politically uncontroversial, and — crucially — the regulated entities (the technology platforms) faced strong public pressure and minimal financial incentive to oppose it. Where political will exists and financial opposition is weak, regulation can move quickly.

AI election regulation does not have the same political economy. The financial incentives against it are substantial, the beneficiaries of the status quo are politically powerful, and the harm is abstract and distributes across the population rather than concentrating in visible victims.

Related reading

Who Lobbied and What They Said

The American Association of Political Consultants, the Internet Association (now renamed NetChoice), and the Computer & Communications Industry Association have all been active in opposing strong federal AI election regulation since the FEC initiated its rulemaking process.

Their public arguments are familiar: regulation would restrict First Amendment-protected political speech; defining “AI-generated” with sufficient precision for regulatory purposes is technically difficult; disclosure requirements impose compliance burdens on small campaigns without addressing the actual harm; platform-level regulation raises government censorship concerns.

These arguments are not fabricated. The First Amendment concerns are genuine. The definitional challenges are real. The compliance burden on small campaigns is a legitimate consideration.

The arguments are also strategically deployed to prevent regulation that would constrain the clients of the entities making the arguments. Political consulting firms that have built AI-enhanced targeting practices have a financial interest in those practices not being regulated. Technology platforms that earn political advertising revenue have a financial interest in political advertising remaining minimally regulated. The arguments that happen to align with these financial interests are the arguments that get funded, staffed, and deployed in the lobbying process.

This is standard interest group politics. It is not unique to AI regulation. But understanding it as standard interest group politics — rather than as good-faith technical objection — is necessary for understanding why the regulatory vacuum persists.

The Politicians’ Dilemma

There is a particular structural problem that makes AI election regulation uniquely difficult to pass through a legislative body: the people who would vote on the regulation are the primary users of the unregulated technology.

Members of Congress run political campaigns. Those campaigns use AI targeting, AI-generated content, and AI-optimized messaging. The regulation being proposed would constrain their campaigns’ use of these tools. Members of Congress considering AI election regulation are not evaluating a policy that affects someone else’s interests. They are evaluating a policy that affects their own political survival.

This conflict of interest is not formally disqualifying — members of Congress vote on tax legislation that affects their personal finances, on defense appropriations that affect the value of their stock portfolios, on regulatory policy that affects their donors’ businesses. The American system does not routinely require recusal for general conflicts of interest. But it does mean that the political instinct toward self-preservation runs directly against the policy substance of strong AI election regulation.

Republican members of Congress who have benefited from AI-powered micro-targeting in competitive primaries are not enthusiastic about rules that would constrain micro-targeting. Democratic members who have used AI-generated content in their digital advertising are not enthusiastic about disclosure requirements that would apply to their campaigns. The partisan disagreements about AI election regulation have real intellectual content, but they also track the partisan distribution of tactical advantages from the status quo.

The Revolving Door Mechanism

The regulatory agencies with jurisdiction over election-related AI — primarily the FEC — have a specific problem beyond the usual challenges of regulatory agencies facing sophisticated regulated industries.

The FEC’s staff includes lawyers and policy experts who move, with some frequency, between government positions and private practice representing campaigns, political parties, and political technology companies. The Federal Election Commission, like most regulatory agencies, relies on the expertise that comes from close familiarity with the regulated industry. That familiarity comes, in part, from having worked in that industry.

This is not a corruption story. The revolving door produces genuine expertise alongside genuine conflicts. The problem is subtler: when the policy experts shaping FEC rulemaking are people who have spent their careers in the political consulting industry and expect to return to it, the rules they write tend to reflect the operational constraints and preferences of that industry. Not because they are dishonest, but because their mental model of what is technically feasible and operationally reasonable is shaped by their professional experience.

The AI companies themselves — Anthropic, OpenAI, Google DeepMind — have a more complex relationship to AI election regulation. They have a business interest in broad AI adoption, which argues against heavy regulation. They also have a reputational interest in not being associated with election manipulation, which argues for some regulation. The corporate positions of major AI companies on election AI regulation have been carefully positioned to support “responsible” regulation while opposing specific provisions that would constrain their commercial customers’ use of their products.

The International Comparison

The gap between what European democracies have done on AI election regulation and what the United States has done is instructive about what is technically and institutionally possible, and what requires political will.

Germany passed an AI Election Integrity Act in September 2025, three months before the federal election it was most immediately relevant to. The law requires disclosure of AI use in political advertising, creates criminal liability for synthetic media depicting candidates without consent, establishes a rapid-response electoral court process for pre-election adjudication of AI disinformation complaints, and imposes significant fines on platforms that fail to promptly remove confirmed synthetic political disinformation.

Related reading

The German law is imperfect. It was passed quickly and shows signs of rushed drafting. Enforcement has been uneven. The foreign operation targeting the federal election ran for months despite the law’s existence. But Germany produced a substantial legislative response to the problem in under two years, with criminal liability provisions and rapid enforcement mechanisms.

The United States has produced FEC rulemaking that began in 2023 and remains uncompleted. The difference is not that Germany is faster at rulemaking than the United States. The difference is that the political economy of German AI election regulation is different: German political parties that would be regulated have a different relationship to the regulatory process, technology companies have less lobbying power in the German context, and the political consequences of being seen as blocking election integrity regulation are different.

The 2026 Status and What Passed

The 2026 legislative session produced two pieces of federal legislation touching on AI in elections.

The Protect Elections from Deceptive AI Act passed the Senate in May 2026 on a bipartisan vote of 64-36. The act prohibits AI-generated video or audio in political advertising that “materially misrepresents” a federal candidate. It includes disclosure requirements for AI use in political advertising and creates FEC enforcement mechanisms. The House has not brought the bill to a floor vote as of October 2026.

A second bill, the Honest Ads Act Extension, which would have extended existing paid political advertising disclosure requirements to cover AI-generated content more comprehensively, passed committee but has not received a floor vote.

The Senate bill’s failure to advance in the House reflects the combination of factors discussed above: technology industry opposition to specific enforcement provisions, Republican concerns about FEC overreach, and the absence of political urgency sufficient to overcome the inertia of a body with many competing legislative priorities.

The state-level patchwork continues to grow. California, Texas, Georgia, Washington, Minnesota, and Florida all have state-level AI election regulations of varying scope. These laws are being challenged in federal court, enforced inconsistently, and exploited for definitional loopholes. They represent genuine regulatory activity and genuine inadequacy in roughly equal measure.

The Counterfactual That Clarifies the Analysis

The clearest way to see the political economy driving the regulatory vacuum is to ask: what would federal AI election regulation look like if it were designed primarily to solve the problem rather than to satisfy the political actors who would constrain it?

It would include: criminal liability for provably false AI-generated political content distributed at significant scale; mandatory cryptographic provenance requirements for political advertising content; platform obligations to detect and label AI-generated political content, backed by substantial fines for non-compliance; a rapid electoral court process for pre-election adjudication of AI content complaints; and comprehensive voter data privacy requirements that would degrade the commercial data ecosystem underlying AI political targeting.

This regulatory architecture is technically feasible. Versions of each element have been implemented elsewhere or proposed in American legislative contexts. The reason this architecture does not exist in the United States is not technical feasibility or regulatory capacity. It is political feasibility — each element encounters substantial opposition from interests powerful enough to block it.

Criminal liability for AI political content: opposed by technology companies (platform liability) and political campaigns (content liability), on First Amendment grounds.

Mandatory provenance for political advertising: opposed by campaigns that want operational flexibility and technology companies that would bear implementation costs.

Platform obligations: opposed by platforms on First Amendment and compliance cost grounds.

Electoral court process: opposed by political actors who benefit from delayed adjudication — the election is decided before the court rules.

Voter data privacy: opposed by commercial data brokers and the political consulting industry that depends on their products.

Related reading

The sum of these oppositions is the regulatory vacuum. Understanding it as a deliberate outcome — the predictable product of specific political economy — rather than as an accident of institutional slowness is the beginning of understanding what it would actually take to change it.

What Change Requires

The historical record of major regulatory change in American democratic governance suggests two mechanisms that have successfully overcome comparable structural opposition.

The first is crisis with clear victims. The Pure Food and Drug Act of 1906 followed muckraking journalism that created clear victims with whom the public identified. The Securities Exchange Act of 1934 followed a financial crisis in which the costs of inadequate regulation were visible and catastrophic. The Americans with Disabilities Act and the Clean Air Act each passed after sustained campaigns that made the harms of the status quo vivid and attributable.

AI election harm is diffuse, hard to attribute, and largely invisible to the voters experiencing it. The Wisconsin business owner, Maria Chen, the German voters misled about Esken — these are stories that got some coverage. They have not produced the kind of broad public identification with victims of AI election harm that historically drives regulatory change against entrenched opposition.

The second mechanism is partisan alignment — cases where both parties’ political interests happened to point toward regulation. Campaign finance reform after Watergate is the clearest example: the scale of the Nixon campaign’s abuses was severe enough that Republican opposition to reform was untenable. Bipartisan regulation emerged when the cost of defending the status quo exceeded the cost of accepting constraints.

AI election regulation has not yet produced a crisis severe enough, or a political realignment severe enough, to overcome the structural opposition. The 2026 election cycle will contribute to the cumulative evidence. Whether it contributes enough to change the political calculus is the key question for the next legislative session.

History suggests change comes when the cost of the problem becomes clearly greater than the cost of the solution to enough political actors. That calculation has not yet tipped. But it is moving.