Ruling Protects Political Speech but Opens Regulatory Pandora’s Box

A federal judge’s decision to strike down Montana’s AI deepfake election law as unconstitutional marks a significant win for free speech advocates – but the narrow scope of the ruling and evolving legal landscape suggest this victory may be a Pyrrhic one, potentially paving the way for more sophisticated restrictions on political expression.

U.S. District Judge Susan Watters ruled this week that Montana’s new law regulating artificial intelligence-generated deepfakes in political campaigns violates the First Amendment’s protection of political speech. The decision came in response to complaints filed against a political action committee that used AI-altered images of candidates in mailers during the June primary election. While civil liberties groups are celebrating the ruling as a triumph for free expression, a closer examination reveals troubling implications that extend far beyond this single case.

The immediate headlines celebrate the preservation of political speech rights, with attorney Adam Schulman of the Hamilton Lincoln Law Institute declaring it “an important victory for political speech and the First Amendment.” However, the decision’s remarkably narrow application – affecting only the named plaintiffs rather than invalidating the law statewide – exposes a more complex and potentially dangerous trajectory for First Amendment protections in the digital age.

The controversy centers on mailers distributed by a PAC led by Dan Bartel, a former Republican legislator, that depicted political candidates including Rep. Eric Albus, former legislator Jennifer Carlson, and Rep. Llew Jones with Pride flags and buttons. The images were created using artificial intelligence to alter the candidates’ appearances, prompting complaints to Montana’s Commissioner of Political Practices under the state’s new AI deepfake law.

Judge Watters determined that the Montana statute engaged in “viewpoint-based discrimination” by treating “identically deceptive media differently based solely on whether its message is weaponized or deployed to elevate a candidate.” This distinction, she argued, made the law “presumptively unconstitutional” under established First Amendment jurisprudence that affords political speech the highest level of protection.

The legal reasoning appears sound on its surface. Political speech has long been considered the most protected form of expression under the Constitution, and regulations that discriminate based on the viewpoint expressed have historically faced the strictest scrutiny from courts. However, the devil lurks in the details of how this precedent will shape future regulatory efforts.

A Montana federal judge blocked enforcement of the state’s AI deepfake election law against a political action committee, finding that the law likely discriminates based on viewpoint by treating similar deceptive media differently depending on whether it helps or attacks a candidate. The preliminary ruling does not invalidate the law statewide or eliminate the government’s ability to regulate synthetic media. Instead, it highlights the constitutional risks of broad restrictions while leaving lawmakers room to pursue narrower rules focused on disclosure, impersonation, fraud, and concrete election-related harms.

What makes this ruling particularly noteworthy – and concerning from a civil liberties perspective – is its extraordinarily limited scope. Judge Watters explicitly stated that her decision applies only to Dan Bartel and his PAC, leaving the law fully enforceable against every other political actor in Montana. This limitation stems from recent Supreme Court jurisprudence, particularly the Trump v. CASA decision, which dramatically curtailed federal courts’ ability to issue broad preliminary injunctions that strike down laws universally.

The Trump v. CASA ruling fundamentally altered how injunctions function in federal courts. The Supreme Court’s conservative majority determined that the Judiciary Act of 1789 does not authorize federal courts to impose universal injunctions that protect parties beyond those directly involved in litigation. Instead, courts must now limit equitable relief to named plaintiffs, ensuring remedies are “no more burdensome than necessary.”

This represents a seismic shift in constitutional litigation strategy. Previously, a successful challenge to an unconstitutional law could protect all similarly situated individuals throughout a state or even nationwide. Now, each person or organization must separately sue and win their own injunction to gain protection – a requirement that creates profound practical and constitutional problems.

The implications are stark: Montana’s AI deepfake law remains on the books and enforceable against everyone except Bartel and his specific PAC. Other political committees, candidates, advocacy groups, or individual citizens who wish to engage in similar political speech using AI-generated content must either risk prosecution under what a federal judge has deemed unconstitutional, or file their own lawsuits seeking individual exemptions.

The public discourse surrounding AI deepfakes in politics has constructed a misleading binary: either we protect free speech and allow unregulated AI manipulation of political imagery, or we safeguard electoral integrity by restricting these technologies. This framing serves the interests of those who would expand government power to regulate political expression while simultaneously providing cover for the judiciary to chip away at universal protections.

Judge Watters’ ruling perpetuates this false dichotomy by focusing narrowly on viewpoint discrimination while leaving untouched the broader question of whether governments should regulate AI-generated political content at all. The decision suggests that Montana’s law failed not because the government lacks authority to police AI use in campaigns, but because it discriminated based on whether the content helped or hurt a candidate. This implies that a more carefully drafted statute – one that applies content-neutral restrictions regardless of viewpoint – might pass constitutional muster.

Indeed, research reveals that state legislatures across the country are responding to concerns about AI-generated political content by crafting increasingly sophisticated regulations. According to background research on state deepfake laws, as of early 2026, various states have enacted measures requiring disclaimers on digitally altered political content and imposing liability on both creators and platforms that distribute such material. These laws attempt to navigate constitutional constraints by focusing on disclosure requirements rather than outright bans, creating a regulatory framework that may survive First Amendment challenges even as it chills political speech.

The danger lies in the regulatory arms race this decision encourages. Rather than deterring government overreach, the ruling provides a roadmap for lawmakers to craft more constitutionally defensible restrictions on political expression. Future laws can avoid the viewpoint discrimination trap by imposing uniform burdens on all AI-generated political content, mandating disclaimers, registration requirements, or pre-clearance procedures that apply regardless of whether the speech supports or opposes a candidate.

The strongest regulatory principle is to target demonstrable harm rather than synthetic media or political viewpoint. Clear, affordable disclosures and carefully defined prohibitions on fraudulent impersonation may protect voters without giving officials broad power to decide which political messages are acceptable. But formal neutrality is not enough: lawmakers and courts must also examine whether compliance costs, vague standards, or severe penalties silence grassroots speakers in practice.

The practical consequences of the Trump v. CASA framework, as applied in this Montana case, create what legal scholars recognize as a profound chilling effect on political speech. When a law remains enforceable against everyone except successful litigants, it transforms constitutional rights from universal protections into privileges available only to those with the resources, sophistication, and willingness to file federal lawsuits.

Consider the position of a small grassroots organization or individual activist who wishes to create political content using AI tools. Even knowing that a federal judge has ruled Montana’s law unconstitutional as applied to similar conduct, that organization or individual faces prosecution risk. They must choose between self-censoring their political expression or undertaking expensive, time-consuming litigation with uncertain outcomes. The very existence of the law, even after being ruled unconstitutional in one application, deters protected speech.

This fragmentation of constitutional protections creates a two-tiered system of rights: well-funded political actors like Bartel’s PAC can afford to litigate their way to freedom, while smaller players must either comply with laws that courts have deemed unconstitutional or risk enforcement actions they cannot afford to defend against. The result is precisely the sort of inequality in political speech that the First Amendment was designed to prevent.

Moreover, the limited scope of injunctions encourages selective enforcement. Political Practices commissioners and prosecutors can continue applying laws that have been ruled unconstitutional, knowing that each new enforcement action requires a separate legal challenge. This dynamic shifts power dramatically toward government regulators and away from citizens exercising their constitutional rights.

The Montana ruling does not exist in isolation but rather represents one data point in a rapidly evolving legal landscape surrounding AI-generated political content. Research indicates that 72% of Americans believe deepfakes will influence elections, and 70% report feeling more skeptical toward online content overall. This public anxiety creates political pressure for lawmakers to “do something” about AI in campaigns, often without careful consideration of constitutional boundaries.

Federal and state legislative proposals increasingly focus on disclosure requirements, platform accountability measures, and criminal penalties for certain uses of AI in political contexts. These regulations, while facially content-neutral, impose significant burdens on political speech and carry enforcement mechanisms that can be selectively deployed against disfavored speakers or viewpoints.

The trajectory appears clear: as AI tools become more sophisticated and accessible, regulatory frameworks will expand and become more nuanced in their attempts to survive constitutional scrutiny. Laws will shift from crude bans toward complex regulatory schemes involving mandatory disclosures, registration requirements, and procedural hurdles that achieve through administrative burden what cannot be accomplished through direct prohibition.

Judge Watters’ decision, by invalidating Montana’s law only on viewpoint discrimination grounds while affirming the government’s general interest in regulating AI deepfakes, effectively invites this legislative refinement. Future statutes that avoid the specific constitutional pitfall identified in this case may prove far more durable and far more restrictive of political expression.

The Montana ruling illustrates a troubling pattern in contemporary First Amendment jurisprudence: courts issue narrow decisions that nominally protect speech rights while simultaneously establishing frameworks that enable more sophisticated forms of government control. The combination of limited injunctive relief and nuanced constitutional analysis creates an environment where rights are honored in theory but become increasingly difficult to exercise in practice.

Political speech advocates face a challenging landscape ahead. Each legislative session brings new proposed restrictions on AI-generated content, each crafted with slightly more constitutional sophistication than the last. Meanwhile, the Supreme Court’s limitation on universal injunctions ensures that even successful constitutional challenges provide limited protection to the broader community of speakers.

The false dichotomy between protecting speech and preventing electoral manipulation continues to drive policy debates, obscuring the reality that overly broad regulations pose their own threat to democratic integrity. When government assumes the power to determine which political speech requires disclaimers, which AI-generated content crosses legal lines, and which uses of technology in campaigns warrant criminal penalties, it acquires dangerous tools for suppressing dissent and controlling democratic discourse.

The Montana decision may be celebrated as a victory for political speech, but it simultaneously opens a Pandora’s box of regulatory possibilities. By invalidating one poorly drafted law while affirming the government’s interest in regulating this space, and by limiting relief to a single PAC while leaving the statute enforceable against everyone else, Judge Watters has created a precedent that may ultimately do more harm than good to First Amendment freedoms.

As artificial intelligence becomes increasingly central to political communication, the stakes of these legal battles will only grow higher. The question is not whether courts will protect political speech in individual cases, but whether the cumulative effect of narrow rulings, limited injunctions, and increasingly sophisticated regulations will create a regime where constitutional protections exist in name only, available primarily to those with the resources to litigate their way to freedom while others must navigate an ever-expanding web of restrictions on their political expression.

The Montana AI deepfake ruling represents both a tactical victory and a strategic warning: in the battle to preserve robust protections for political speech in the age of artificial intelligence, winning individual cases may matter far less than the broader regulatory and procedural frameworks those victories help establish.