AI Deepfakes Put New Pressure on Election Trust

Published: 30 September 2026. The English Chronicle Desk. The English Chronicle Online

The rapid spread of artificial intelligence-generated images is creating a new challenge for democratic politics, as fabricated photographs and other synthetic media become increasingly difficult for ordinary voters to distinguish from authentic material. The problem is no longer limited to futuristic concerns about technology. Politicians, election officials, campaign organisations and voters are already confronting examples in which apparently convincing images are circulated online before their authenticity can be established.

The experience of College Park, Georgia mayor Bianca Motley Broom illustrates how quickly such material can travel. An artificially manipulated image appearing to show her kissing a man who was not her husband circulated widely on Facebook. Although the image was fake, awareness of that fact developed only after the material had already reached thousands of people.

For Motley Broom, the incident was more than an embarrassing piece of fabricated content. She said it reflected wider problems in political discourse and could reinforce stereotypes about women in leadership. The episode also demonstrated how the barrier to producing a convincing political fake has changed dramatically.

In the past, creating a fabricated photograph of a public figure often required at least some knowledge of professional image-editing software. Generative artificial intelligence has made the process considerably easier. A user can now describe an imagined scene in ordinary language and obtain an image that may look sufficiently realistic to persuade people who encounter it without additional context.

That development has raised particular concern because social media rewards material that attracts immediate attention. A sensational image can be copied, reposted and discussed thousands of times before fact-checkers, journalists or the person depicted have an opportunity to respond.

Public concern about the issue is substantial. A survey conducted in June by the Rad Campaign and Women Who Tech found that about 71% of respondents were worried about AI-generated deepfakes and disinformation affecting elections. Roughly two-thirds also expressed concern about artificial intelligence being used to sexualise women.

Those concerns extend beyond election manipulation. Synthetic images can be used to attack reputations, create misleading impressions about public figures or produce sexualised material involving people who never appeared in the depicted circumstances. The speed and accessibility of generative tools means that the technology can be used by individuals with little technical expertise.

The political implications are particularly serious because elections depend heavily on voters being able to assess competing claims. A fabricated photograph or video does not necessarily need to convince everyone to have an effect. It may simply create enough uncertainty to make voters question genuine information or distrust material they would otherwise consider credible.

That is one reason politicians from different sides of the US political spectrum have been discussing possible responses. Senator Raphael Warnock of Georgia has publicly described encountering fabricated AI images portraying him in circumstances that never happened. Some of the images mocked his appearance, while others falsely depicted him as having health problems or grandchildren.

Warnock has warned that the increasing sophistication of such technology could make the distinction between genuine and fabricated information increasingly difficult. He has also raised the possibility of a coordinated campaign emerging shortly before an election, when there may be little time for journalists, officials or candidates to establish the facts.

The concern is particularly relevant to the 2026 US midterm election environment, where election officials are already preparing for misinformation, cybersecurity threats and disputes over election administration. Recent reporting has found election authorities strengthening preparations for possible disruption and expanding efforts to communicate directly with voters as misinformation circulates through social media and other online platforms.

The US Senate and state legislatures are consequently considering different approaches. One proposed federal response discussed by lawmakers has involved requiring stronger safeguards from artificial intelligence companies. At the state level, legislators have considered requirements for political campaigns to disclose when advertisements have been manipulated using AI.

Tennessee provides one example of how such rules are developing. The state enacted legislation requiring political advertisements containing certain deepfake elements depicting candidates in actions or speech that did not actually occur to carry a disclaimer. The law also provides legal remedies for candidates affected by violations.

Tennessee lawmakers have also considered more specific restrictions on electoral deepfakes. A 2026 proposal sought to restrict deceptive synthetic media depicting candidates or political parties during the 90 days before an election, while allowing exceptions where appropriate disclosures are included. The proposal also contemplated legal remedies for candidates or parties targeted by such material.

The legislative debate, however, is not simply about stopping false information. It also involves constitutional protections for political speech. Any regulation aimed at manipulated political material must contend with the First Amendment and the possibility that broad restrictions could affect legitimate satire, commentary or political expression.

That tension has already appeared in court. In September 2026, a federal judge in Montana temporarily blocked enforcement of a state law targeting certain election-related deepfakes in a case involving a conservative political action committee. The ruling reflected concerns that the law could raise First Amendment problems by treating some political uses of AI-generated material differently according to viewpoint.

Georgia presents its own complicated situation. Lawmakers have previously debated measures dealing with AI-generated campaign material and disclaimers, amid concerns about the effect of synthetic media on the state’s elections. Legislative records have identified AI-generated media as a potential threat to free and fair elections while also recognising the importance of protecting freedom of speech.

The state has therefore become part of a broader national debate over where the line should be drawn between political expression and deliberately deceptive synthetic media. The challenge becomes especially difficult when an image is presented as satire but is subsequently shared without its original context.

The legal debate is also expanding beyond elections. Tennessee’s ELVIS Act, enacted in 2024, expanded protections for an individual’s voice against unauthorised AI-generated cloning, reflecting concerns about synthetic impersonation more broadly.

For people targeted by fabricated political images, however, legislation alone may not solve the problem. Once false material reaches social media, removing the original post does not necessarily remove the copies. Screenshots, reposts and edited versions can continue circulating after the initial claim has been disproved.

This creates an unusual problem for democratic communication. Correcting false information can require considerably more explanation than creating it. A fabricated image may take seconds to generate, while establishing its origin, demonstrating manipulation and communicating the correction can take hours or days.

The problem can also affect genuine information. If voters become accustomed to seeing convincing synthetic photographs and videos, they may begin to distrust authentic material simply because it looks unusual or because someone claims it was generated by AI. In that environment, misinformation can work not only by persuading people to believe something false but also by making reliable information appear uncertain.

For candidates such as Motley Broom, the personal consequences are immediate. A fabricated image can affect family relationships, professional reputations and public perceptions long before a correction reaches the same audience.

She has indicated that legal action could potentially be considered under existing defamation protections, while also stressing that any new legislation should carefully balance individual rights with constitutional protections for speech. Her concern is that false attacks should have consequences without creating rules that unnecessarily restrict legitimate expression.

That balance is likely to remain at the centre of the debate as AI-generated media becomes more sophisticated. Election authorities, technology companies, lawmakers, journalists and voters all face the same underlying problem: determining authenticity at a time when visual evidence can no longer automatically be treated as proof.

The future of election information may therefore depend not only on better detection technology, but also on stronger public habits around verification. A photograph that appears convincing may no longer be enough. The circumstances in which it was published, the account that first shared it, independent confirmation and the availability of original material may all become increasingly important.

The question raised by Warnock — how voters will know what is true when fabricated images and stories can be produced moments before an election — is consequently becoming a practical issue rather than a distant technological warning. As artificial intelligence continues to lower the cost of producing convincing synthetic media, the ability to establish trust in political information could become one of the defining challenges facing modern elections.