AI and election interference

Generative AI has rapidly changed how information is created and disseminated. The technology has created new opportunities for actors intent on misleading, manipulating or undermining confidence in democratic processes and elections.

Advancements in generative AI have lowered the threshold for producing credible images, videos and voice recordings. What used to require advanced technology can now be created by individual users in seconds. The technology is increasingly used to sway opinions and political processes, as witnessed in other countries’ elections.

AI-generated material can include fabricated statements, manipulated videos, or images that show events that never took place. When such content is disseminated on social media, it can rapidly gain a broad reach and influence people’s perceptions.

How AI can be used to sway elections

AI technology enables creating credible yet partly or wholly fabricated images, videos and voice recordings that can be used to the detriment of political candidates or to create distrust in the electoral process itself. By analysing large volumes of data, AI can also be used to direct messages at specific target groups, which risks amplifying polarisation and giving different groups completely different versions of reality.

At the same time, AI can be used to quickly build up fake news sites and accounts that purport to be independent sources of information, but that are in fact run by actors with underlying antagonistic interests. Technological developments also enable producing large volumes of disinformation, lowering the threshold for malign information influence.

With the help of AI, threat actors can create large numbers of fake accounts that resemble real people. These accounts may have profile photos, credible background stories, and a history of posts built up over time. They can take part in discussions, form relationships, and spread messages in ways that mimic human behaviour.

When many such accounts work together, they can generate large amounts of varied content, present different viewpoints, and create the impression that many people share the same opinions. This makes manipulation harder to detect and increases the risk that people’s perceptions and decisions are influenced.

The Psychological Defence Agency considers that foreign actors, especially Russia, are increasingly using AI‑generated material in their information influence targeting Europe.

AI images make fake information more convincing

Studies show that repeated exposure to information causes it to be more frequently perceived as true, even when it may be false. This is because the brain interprets recognition as truth, and repeated exposure to information combined with realistic imagery therefore makes us more vulnerable.

When a lie is paired with a credible-looking image, it is more difficult to distinguish between disinformation and truth. Realistic AI-generated images can be used as evidence and give a sense of credibility to texts that might otherwise be understood to be fake. For example, fake news headlines are perceived to be more credible when accompanied by a realistic image that appears to confirm what the headline is claiming.

Today’s information environment contains a surging volume of “AI slop” – that is, low-quality AI-generated material that is not created to look authentic – the risk of influence has increased. This is because today it is easier, cheaper and more effective for a threat actor to to produce content that does not look entirely realistic. When production becomes cheaper, more content can be created and spread, meaning people encounter it more frequently. Even if the material does not appear particularly credible, repeated exposure can make it feel familiar and therefore seem true.

As it becomes harder to determine what is genuine, people may gradually become more suspicious of images and videos. If it is no longer clear what can be trusted, it becomes more difficult to discuss societal issues, as people form different views about what has actually happened. This can also reduce trust in public authorities and other key institutions, and over time affect how democracy functions.

Risks to be aware of

AI and social media can influence elections in several ways:

• Erroneous information can be spread faster through algorithms that amplify polarising content.
• Fabricated situations can be shared as if they were genuine and risk undermining confidence in the election.
• AI‑generated websites may purport to be independent, but can have ties to a state apparatus or other actors.
• AI‑services can give incorrect answers about candidates or the electoral process, which risks spreading disinformation.