Artificial intelligence and elections

Generative AI capabilities allow creation of misleading content. Examples of this include text-to-video, deepfake videos, text-to-image, AI-altered image, text-to-speech, voice cloning, and text-to-text. In the context of an election, a deepfake video of a candidate may propagate information that the candidate does not endorse.^[2] Chatbots could spread misinformation related to election locations, times or voting methods. In contrast to malicious actors in the past, these techniques require little technical skill and can spread rapidly.^[3]

During the 2023 Argentine primary elections, Javier Milei's team distributed AI generated images including a fabricated image of his rival Sergio Massa and drew 3 million views.^[4] The team also created an unofficial Instagram account entitled "AI for the Homeland."^[4] Sergio Massa's team also distributed AI generated images and videos.^{[5] [6]}

In the run up to the 2024 Bangladesh elections, a strict muslim country, deepfake videos of female opposition politicians appeared.^[7] Rumin Farhana was pictured in a bikini while Nipun Ray was shown in a swimming pool.^[7]

In the 2024 French elections, deepfake videos appeared claiming: i) They showed the family of Marine le Pen. In the videos, young women, supposedly Le Pen's nieces, are seen skiing, dancing and at the beach "while making fun of France’s racial minorities": However, the family members don't exist. On social media there were over 2 million views.^[8] ii) In a video seen on social media, a deepfake video of a France24 broadcast appeared to report that the Ukrainian leadership had "tried to lure French president Emmanuel Macron to Ukraine to assassinate him and then blame his death on Russia".^[9]

In the 2024 Indian general election, politicians used deepfakes in their campaign materials. These deepfakes included politicians who had died prior to the election. Mathuvel Karunanidhi's party posted with his likeness even though he had died 2018.^[10][11][12] A video The All-India Anna Dravidian Progressive Federation party posted showed an audio clip of Jayaram Jayalalithaa even though she had died in 2016.^[13][14] The Deepfakes Analysis Unit (DAU) is an open source platform created in March 2024 for the public to share misleading content and assess if it had been AI-generated.^[15]

AI was also used to translate political speeches in real time.^[11] This translating ability was widely used to reach more voters.^[11][12]

In May 2023, ahead of the New Zealand October 2023 general election, the New Zealand National Party published a "series of AI-generated political advertisements" on its Instagram account.^[16] After confirming that the images were faked, a party spokesperson said that it was "an innovative way to drive our social media".^[16]

AI has been used by the imprisoned ex-Prime Minister Imran Khan and his media team in the Pakistani elections of 2024:^[17] i) An AI generated audio of his voice was added to a video clip and was broadcast at a virtual rally.^[17] ii) An op-ed in the The Economist written by Khan was later claimed by himself to have been written by AI which was later denied by his team.^[17] The article was liked and shared on social media by thousands of users.

In the South African general elections, there were several uses of AI content:^[18] i) A deepfaked video of Joe Biden emerged on social media showing him saying that "The U.S. would place sanctions on SA and declare it an enemy state if the African National Congress (ANC) won".^[18] ii) In a deepfake video, Donald Trump was shown endorsing the uMkhonto weSizwe party. It was posted to social media and was viewed more than 158,000 times.^[18] iii) Less than 3 months before the elections, a deepfake video showed U.S. rapper Eminem endorsing the Economic Freedom Fighters party while criticizing the ANC. The deepfake was viewed on social media more than 173,000 times.^[18]

A committee for one presidential candidate Yoon Suk Yeol released an AI avatar 'Al Yoon Seok-yeol' that would campaign in places the candidate could not go. The other presidential candidate Lee Jae-myung introduced a chatbot that provided information about the candidate's pledges.^[19]

Deepfakes were used to spread misinformation before South Korea's parliamentary elections with one source reporting 129 deepfake violations of election laws within a two week period.^[20] Seoul hosted the 2024 Summit for Democracy, a virtual gathering of world leaders initiated by US President Joe Biden in 2021.^[21] The focus of the summit was on digital threats to democracy including artificial intelligence and deepfakes.^[22]

Steve Endacott created "AI Steve," an AI avatar as the face of his campaign for member of parliament.^[23] The Centre for Emerging Technology and Security provided a report on the threat of AI to the 2024 UK general elections. The report findings said that AI impact was limited but may damage the democratic system.^[24]

Regulation of AI with regard to elections was unlikely to see a resolution for most of the 2024 United States general election season.^[25][26] The campaign for the 2024 Republican nominee,^[27] Donald Trump, has used deepfake videos of political opponents in campaign ads and fake images showing Trump with black supporters.^[25][28] In 2023, while he was still running for re-election, the presidential campaign of Joe Biden prepared a task force to respond to AI images and videos.^[29]

A Democratic consultant working for Dean Phillips also admitted to using AI to generate a robocall which used Joe Biden's voice to discourage voter participation.^[30]

Generative AI increased the efficiency with which political candidates were able to raise money by analyzing donor data and identifying possible donors and target audiences.^[31][32]

The Commission on Elections (COMELEC) issued guidelines on the usage of AI, to be implemented starting from the 2025 Philippine general election including the parallel Bangsamoro Parliament election. It mandates candidate to disclose usage of AI in their campaign materials and prohibits the usage of the technology to spread misinformation against their rivals.^[33] This is the first time the COMELEC has release guidelines on campaigning through social media.^[34]

US states have attempted regulation of AI use in elections and campaigns with varying degrees of success.^[35] The National Conference of State Legislatures has compiled a list of legislation regarding AI use by state as of 2024, some carrying both civil and criminal penalties.^[36] Oregon Senate Bill 1571 requires that campaign communications in Oregon disclose the use of AI.^[37][38][39] California has enacted legislation that makes using deepfakes to discredit political opponents illegal within sixty days of an election.^[40][41]

Midjourney, an AI image-generator, has started blocking users from creating fake images of the 2024 US Presidential candidates.^[42] Research from the Center for Countering Digital Hate found that image generators such as Midjourney, ChatGPT Plus, DreamStudio, and Microsoft's Image Creator create images that constitute election disinformation in 41% of the test text prompts they tried.^[42] OpenAI implemented policies to counter election misinformation such as adding digital credentials to image origin and a classifier to detect if images were AI generated.^[43]

AI has begun to be used in election interference by foreign governments.^[44][45][46] Governments thought to be using AI to interfere in external elections include Russia, Iran and China.^[44] Russia was thought to be the most prolific nation targeting the 2024 presidential election with their influencing operations "spreading synthetic images, video, audio and text online", according to U.S intelligence officials.^[44] Iran has reportedly generated fake social media posts stories and targeted "across the political spectrum on polarizing issues during the presidential election".^[44] China has used "broader influence operations" that aim to make a global image and "amplify divisive topics in the U.S. such as drug use, immigration, and abortion".^[44]

Outside of the US elections, a deepfake video of Moldova’s pro-Western president Maia Sandu shows her "throwing her support behind a political party friendly to Russia."^[45] Officials in Moldova "believe the Russian government is behind the activity".^[45] Slovakia's liberal party leader had audio clips faked which discussed "vote rigging and raising the price of beer".^[45] China has used AI to stir concerns about US interference in Taiwan.^[45] A fake clip seen on social media showed a fake video of the vice chairman of the U.S. House Armed Services Committee promising "stronger U.S. military support for Taiwan if the incumbent party’s candidates were elected in January".^[45]

As the use of AI and its associated tools in political campaigning and messaging increases, many ethical concerns have been raised.^[47] Campaigns have used AI in a number of ways, including speech writing, fundraising, voter behaviour prediction, fake robocalls and the generation of fake news.^[47] At the moment there are no US federal rules when it comes to using AI in campaigning and so its use can undermine public trust.^[47] Yet according to one expert: "A lot of the questions we're asking about AI are the same questions we've asked about rhetoric and persuasion for thousands of years."^[47]

As more insight into how AI is used becomes ever greater, concerns have become much broader than just the generating of misinformation or fake news.^[48] Its use by politicians and political parties for "purposes that are not overtly malicious" can also raise ethical worries.^[48] For instance, the use of 'softfakes' have become more common.^[48] These can be images, videos or audio clips that have been edited, often by campaign teams, "to make a political candidate seem more appealing."^[48] An example can be found in Indonesia's presidential election where the winning candidate created and promoted cartoonish avatars so as to rebrand himself.^[48]

How citizens come by information has been increasingly impacted by AI, especially through online platforms and social media.^[49] These platforms are part of complex and opaque systems which can result in a "significant impact on freedom of expression", with the generalisation of AI in campaigns also creating huge pressures on "voters’ mental security".^[49] As the frequency of AI use in political campaigning becomes common, together with globalization, more 'universalized' content can be used so that territorial boundaries matter less.^[49] While AI collides with the reasoning processes of people, the creation of "dangerous behaviours" can happen which disrupt important levels of society and nation states.^[49]

In the future it is likely that AI will further revolutionise campaigning by the use of, for example, speech analysis and policy development.^[32] Campaign assistants could interact with voters by answering questions which provides information about policies leading to greater accessibility.^[32] The cyber security of political campaigns could be enhanced by protecting data from malicious hackers.^[32] As technologies advance, AI could create more innovative tools for campaigning, which might also bring "challenges in terms of ethical use and data privacy".^[32] Over time, AI-powered political campaigns are likely to become more "data-driven, efficient, and tailored to the evolving dynamics of voter behavior and preferences".^[32]

• List of elections in 2025
• Use of artificial intelligence - AI used in Trump 2024 presidential campaign.

• "Smashing Security: Keeping the lights on after a ransomware attack" - podcast including discussion on the use of AI in the Indian elections (17m37s - 29m11s). 25 April 2024.