Generative AI (genAI) poses a number of risks to elections, such as amplifying disinformation, facilitating foreign interference, and automating voter suppression campaigns. All of these capabilities have been utilized in elections around the world in recent years, including in the U.S. Some instances include a genAI robocall purporting to be President Biden urging Americans not to vote during a primary, a genAI campaign aimed at manipulating pro-Ukraine Americans, and Russian attempts to interfere in the U.S. election using genAI. But many threats that were postulated before the 2024 elections did not appear to manifest in large volumes, and ultimately the potential threats from genAI did not impact the outcomes of U.S. elections. Still, the risks from genAI are real and affect a variety of election stakeholders: election officials, candidates and campaigns, voters, government institutions, and technology companies.
Among those stakeholders, genAI poses unique opportunities and challenges for political campaigns. Campaigns can be targets of deepfakes (defined as images, videos, or audio depicting someone doing or saying something that they did not do or say), gen-AI developed phishing campaigns, and spoof websites. Campaigns also stand to benefit from AI tools, which can help create campaign-related content, translate materials, analyze data, and generally act as a force multiplier across many campaign operations.
In 2024, campaigns had a very short time to navigate the rise of this technology, both to identify and mitigate risks from these new tools and to develop skills to utilize them. Given the opportunities and the risks, it is noteworthy that most misuse did not come from campaigns themselves. However, that doesn’t mean that they will remain constrained in future cycles starting with the upcoming 2026 midterm elections, even while they continue to be impacted by misuse by other actors.
Prior research suggests that political campaigns already use genAI to accelerate data analysis and create campaign materials. Given that broad access to the technology is still relatively new, use cases are likely to advance substantially as adoption and access increases. In 2024, some campaigns used genAI in public-facing and even provocative ways: for example, three campaigns for President — Francis Suarez, Dean Phillips, and Robert F. Kennedy Jr. — used genAI to present voters with chatbot avatars of their candidate (the latter two were suspended by OpenAI, though Kennedy’s returned after switching model providers). Shamaine Daniels, a candidate for Pennsylvania’s state legislature, deployed a fully synthetic robocaller. None of these campaigns was successful, but all received widespread media coverage for their efforts.
This report examines the use of genAI during U.S. campaigns, with an eye toward potential harms in upcoming elections beginning with the 2026 midterms. It documents how political campaigns used these technologies in 2024 in order to help policymakers and government agencies, technology companies, and political campaigns to prepare for future elections.
Scope and Methodology
This report relies upon 20 interviews with United States political consultants and political genAI vendors on both sides of the political aisle, tech industry professionals, and other relevant stakeholders. Interviews were conducted from October to December of 2024, with ten conducted before the 2024 U.S. election and ten following the election. While many individuals agreed to be named, some chose to remain anonymous in order to speak more freely. Participants provided insights into current use cases of genAI, the barriers they have encountered, suggestions for future guardrails, and potential risks of ill-conceived or overly broad regulation.
For the purposes of this report, generative artificial intelligence (genAI) is defined as a form of machine learning, often but not exclusively built upon large language models (LLMs), capable of creating systems that create original content in the form of text, audio, image, or video in response to user prompts by identifying and drawing on patterns in enormous training datasets.
While genAI itself is not new, OpenAI’s release of ChatGPT in November 2022 marked a turning point in public access to the technology. Participants in this research noted that use of genAI in politics is still nascent, yet global adoption of the technology spiked in 2024 in both corporate and personal contexts. GenAI use especially grew in the areas of marketing, sales, and product development.