The question that kept election security experts awake through 2024 and 2025 — whether artificial intelligence deepfakes could meaningfully influence democratic elections — has now been definitively answered. The answer is yes. And in 2026, with major elections scheduled across the United States, the United Kingdom, Scotland, France and dozens of other countries, the threat has never been more acute or more sophisticated.
Deepfakes — synthetic audio, video and image content generated by artificial intelligence — have made a journey in less than three years from research curiosity to operational weapon. What once required specialised technical expertise and significant computing resources can now be produced in minutes using cheap cloud-based AI models, localised into dozens of languages and distributed through social media algorithms before any fact-checker or election authority has had the chance to respond.
The data is alarming. Recorded Future, one of the world’s leading threat intelligence firms, documented 82 high-profile deepfake impersonations across 38 countries in the twelve months between July 2023 and July 2024 alone. Electoral manipulation accounted for 15.8 percent of those cases — a proportion that analysts say significantly underestimates the true scale because the majority of deepfake political content is never detected, let alone reported.
A survey of American voters conducted in early 2026 found that 58 percent of US adults expect AI-generated disinformation to escalate before the midterm elections later this year. Federal regulators have publicly acknowledged they are struggling to address the volume of AI-generated content flooding the campaign season.
What 2025 Proved About the Deepfake Threat
The incidents of 2025 provided a series of disturbing proofs of concept that removed any remaining doubt about the practical danger of election-targeted deepfakes.
In Ireland, a fabricated broadcast that appeared to originate from RTÉ — the country’s national public broadcaster — falsely claimed that the presidential election had been cancelled. The clip gathered thousands of shares on Facebook and YouTube before the platforms removed it. The damage — confusion, suppressed turnout in some areas, eroded trust in official communications — had already been done.
In the United States, Fox News acknowledged in November 2025 that it had reported as factual news a story that was based on an AI-generated deepfake video. One of the most-watched news networks in the country had been deceived by synthetic content — and had broadcast that deception to millions of viewers before the error was identified.
These were not isolated incidents in fringe information environments. They were failures in mainstream media and established social platforms — the institutions that democratic societies depend on to maintain a shared factual baseline.
The March 2026 Rycroft Report, an independent review commissioned by the UK government into foreign financial influence and interference in British politics, was unusually blunt in its conclusions. Britain, the report stated, is already experiencing information warfare. Its defences are worryingly weak.
How Foreign State Actors Are Weaponising Deepfakes
The most sophisticated and systematic use of AI deepfakes for electoral interference is not coming from domestic political actors — it is coming from foreign state actors with strategic interests in the outcomes of democratic elections in other countries.
Russia has been the most extensively documented actor in this space. In Moldova, pro-Western President Maia Sandu has been a persistent target of Russian-origin deepfake campaigns. Fabricated videos depicting her endorsing Russian-friendly political parties and announcing plans to resign from office circulated repeatedly ahead of local elections. Moldovan intelligence officials have attributed the campaigns directly to Russian state actors seeking to undermine the country’s westward orientation and damage trust in its democratic institutions.
Iran has been documented running AI-enhanced social engineering and deepfake video creation campaigns specifically targeting American public opinion, according to data from the Center for Foreign Interference Research. The April 2026 documentation of these campaigns coincided with the period of heightened US-Iran nuclear tensions — suggesting a direct connection between geopolitical objectives and information warfare operations.
China has deployed deepfake technology in Taiwan, a self-governing island that Beijing claims as its own territory. A deepfake video showing a senior US congressional representative making promises about American military support for Taiwan was distributed through TikTok ahead of the island’s elections — a sophisticated operation that simultaneously sought to influence Taiwanese voters and inflame tensions between Taiwan and the United States.
The pattern across these operations is consistent. Foreign state actors use deepfakes not primarily to change individual votes but to erode the foundational trust that democratic societies require to function — trust in media, trust in institutions, trust in the authenticity of what political leaders say and do.
The Speed Problem That Governments Cannot Solve
What makes AI deepfakes uniquely dangerous as an electoral weapon is not their sophistication — it is their speed. Traditional disinformation campaigns required time, resources and distribution networks. A well-crafted deepfake video can be produced, localised into multiple languages and distributed to millions of people in the time it takes a fact-checking organisation to begin its verification process.
Election authorities in the United States, India and across the European Union have all confirmed active investigations into AI-generated political content. But investigation after the fact addresses the legal accountability question — it does not address the harm already done to voters who encountered the content before it was removed.
Research consistently demonstrates that corrections and debunking have a limited effect on beliefs formed through initial exposure to false content. A voter who watches a convincing deepfake of a candidate making a shocking statement carries that impression even after being told the video was fabricated. The psychological impact of deepfake exposure is not fully reversible.
Governments and platforms are racing to establish detection standards and rapid response mechanisms. But AI detection technology is in a permanent arms race with AI generation technology — and the generation side currently holds the advantage.
The Legal Framework That Doesn’t Exist Yet
Perhaps the most damaging aspect of the deepfake threat to democratic elections is legal: most election laws in most democracies were written before generative AI existed and are comprehensively ill-equipped to address synthetic political media.
In the United States, federal lawmakers have been debating disclosure requirements for AI-generated political advertisements since 2024 but have not passed comprehensive legislation. The patchwork of state-level laws — some states have banned political deepfakes outright, others have no relevant legislation at all — creates a regulatory landscape that sophisticated actors can exploit through jurisdictional arbitrage.
The European Union is the closest any major democratic bloc has come to a comprehensive regulatory framework. The Digital Services Act creates obligations around the labelling and removal of synthetic content on large platforms. The EU’s AI Act, which began phasing in during 2025, includes specific provisions related to AI-generated content that could affect democratic processes. But enforcement across 27 member states with varying technical capabilities and political priorities remains deeply uneven.
India’s election authorities have taken an aggressive posture, demanding that platforms remove manipulated political media immediately and preserve evidence for potential prosecution. But India’s scale — over 900 million registered voters, elections conducted across dozens of states in multiple languages — makes systematic monitoring effectively impossible with current technology.
What Platforms Are and Are Not Doing
The major social media platforms have each made commitments to address AI-generated election content, but the gap between commitment and implementation remains significant.
Meta has introduced AI content labelling policies that require political advertisers to disclose when their content was created or substantially modified using AI. Google has implemented similar requirements for political ads on YouTube and Search. TikTok has committed to labelling deepfakes and has invested in detection technology. X, formerly Twitter, has largely relied on community notes and user-generated fact-checking — an approach that critics argue is wholly inadequate for the speed and scale of the deepfake threat.
The fundamental problem is that platform policies are enforced at the point of upload — but most political deepfakes do not begin their life as clearly political content on major platforms. They circulate first through messaging apps, private groups and obscure channels before reaching mainstream platforms, by which point they have often already achieved significant reach.
Content watermarking and digital provenance tools — technologies that cryptographically embed information about a video’s origin and editing history into the file itself — represent the most technically promising long-term solution. Several governments and major news organisations have begun experimenting with standards including Content Credentials, developed by the Coalition for Content Provenance and Authenticity. But adoption remains limited and the technology is not yet deployed at the scale required to make a meaningful difference in 2026 election cycles.
Scotland’s Warning
Scotland provides a particularly current case study in the challenges facing election authorities. With a Scottish Parliament election scheduled for May 2026, researchers at the Scottish Election Study sought to measure how aware Scottish voters are of the deepfake threat and how worried they are about it.
The findings were concerning. Awareness of AI deepfakes as an electoral threat was lower than researchers had anticipated. Confidence in voters’ ability to identify synthetic content was higher than their actual detection rates — a dangerous combination. Voters who felt confident in their ability to spot deepfakes were, counterintuitively, more vulnerable to manipulation because their confidence reduced their critical scrutiny of the content they encountered.
The Rycroft Report’s conclusion — that Britain is already in information warfare and its defences are worryingly weak — applies with equal force to Scotland’s election environment.
The Deeper Crisis
The deepfake threat to elections is ultimately a threat to something more fundamental than any individual electoral outcome. It is a threat to the epistemic foundations of democratic governance.
Democracy requires that citizens share a common factual reality from which they can make meaningful choices. When synthetic media makes it impossible to determine whether a video of a candidate is real or fabricated, when the leader of a country cannot be certain that their image and voice are not being used to endorse positions they have never held, when election results become matters of contested perception rather than established fact — the basic functioning of democratic institutions becomes imperilled.
Artificial intelligence did not create the crisis of disinformation that democracies are navigating in 2026. But it has given that crisis a velocity and a scale that existing legal, technical and institutional frameworks were not designed to handle.
The race between the technology that creates deepfakes and the technology that detects them, between the actors who deploy them and the regulators who seek to constrain them, between the damage they do and the corrections that follow — that race will define the integrity of democratic elections for the decade ahead.
GlobeBuzz will continue tracking AI election interference developments worldwide.
Follow @globebuzz_org on X and Telegram for live updates.
