AI and Elections: The Policy Debate Over Political Deepfakes and Disclosure

Elections have always attracted misinformation, but AI-generated political content adds a genuinely new dimension to a long-standing, familiar problem: convincing fabricated audio and video of candidates saying or doing things that never actually happened, now producible faster, cheaper, and more convincingly than ever before. Policymakers worldwide are actively working through how to address this without undermining legitimate political speech and expression in the process.

Why Political Deepfakes Pose a Genuinely Distinct Risk

Political deepfakes carry particular risk because they can spread rapidly during the specific, narrow window of an active election cycle, when there may not be sufficient time for a thorough fact-check or an effective, widely seen correction to actually catch up with and meaningfully counter the original, false, viral claim before people have already voted based partly on that mistaken impression. A convincing fabricated video released close enough to an election could plausibly influence real voting behavior before it’s credibly and widely enough debunked to actually matter.

The Disclosure Requirement Approach

A number of jurisdictions have moved toward requiring clear disclosure whenever AI-generated content appears in political advertising, mandating that such content be clearly and conspicuously labeled as AI-generated or synthetically manipulated in some way. This regulatory approach aims to preserve the ability to actually use these tools in legitimate political contexts while ensuring viewers are actually informed enough to weigh the content appropriately, rather than banning the underlying technology’s political use outright.

Why Outright Bans Face Real Practical Difficulty

More aggressive proposals to outright ban AI-generated political content entirely face real, well-recognized practical and legal difficulty, including genuine challenges in reliably defining exactly what counts as prohibited “AI-generated” content as these tools become more deeply and routinely embedded in completely ordinary content production and editing workflows, and legitimate free-speech concerns in jurisdictions with strong constitutional protections for political speech and expression specifically.

The Cross-Border Enforcement Problem

Election-related disinformation frequently originates from outside the specific jurisdiction actually holding the election in question, creating genuine practical enforcement challenges for any purely domestic regulatory approach. A country can mandate AI-content disclosure for its own domestic political advertisers, but that same domestic mandate has essentially no direct legal reach over foreign actors deliberately producing and spreading deceptive political deepfake content from entirely outside that jurisdiction’s borders.

The Platform Responsibility Question

A significant, actively debated part of this broader policy conversation concerns what specific responsibility social media and other content platforms themselves should bear for detecting and appropriately labeling or removing AI-generated political disinformation circulating on their own services, an obligation that requires real platform investment in detection capability and, more fundamentally, requires platforms to make often genuinely difficult, contested judgment calls about content that sits in a real gray area rather than being unambiguously, obviously fabricated.

Why Timing Makes This Regulatory Area Especially Hard

Election-related AI regulation faces an especially difficult and somewhat unusual timing challenge: rules genuinely need to be in place and well understood well before an actual election takes place, but the underlying technology itself, and the specific tactics bad actors use to exploit it, keep evolving quickly enough that rules crafted even a relatively short time in advance can risk becoming outdated or leaving new, unanticipated gaps by the time an election actually happens.

What Voters Can Reasonably Do in the Meantime

While policy frameworks continue to develop and mature, voters encountering particularly surprising, dramatic, or inflammatory political audio or video, especially close to an actual election, benefit from applying real extra scrutiny — checking whether credible, independent news organizations have separately corroborated the specific content, and being appropriately cautious about content that arrives through unofficial, unverified channels with no clear, traceable, credible original source behind it.

Further reading: See our guides to Deepfakes Explained: How They’re Made and How They’re Detected and The State of AI Regulation: What the EU AI Act Means for Businesses and Developers. For background, see Wikipedia’s entry on Deepfake.

Want to contribute to Daily AI News? We welcome guest posts, research write-ups, and press releases from the AI community. Write for us →