Deepfakes in Electoral Campaigns

State Lawmakers Address A Direct Threat from AI: Their Own Electoral Campaigns

As lawmakers contemplate their reelection campaigns, a direct concern of theirs, a concern shared by industry leaders, is AI’s use to influence future election results. AI, in the form of “deepfake” images, audio, and video could be used to manipulate images and recordings to change a candidate’s speech into something they didn’t say, alter a candidate's movement in an embarrassing manner, or doctor images to perpetuate false narratives. And in today’s viral social media environment, an order of magnitude more voters will view and share these deepfakes than will see any fact-check posts revealing the truth.

Deceptive media produced by AI to influence elections may originate from a variety of sources — individual campaigns, outside groups seeking to oppose or support a candidate or issue, or foreign adversaries seeking to influence the outcome of an election. It’s unclear at this point what type of impact deceptive AI deepfakes could have on elections, but this potential impact is being taken especially seriously given social media’s influence on recent elections. A repeated theme we’ve seen from policymakers looking to regulate AI is the regret that lawmakers failed to act on social media and that they won’t make the same mistake with AI.

And we’re already seeing the early effect of media distortions on elections. In 2020, a manipulated video of former House Speaker Nancy Pelosi went viral, which artificially slowed down her speech and made it appear slurred, falsely indicating that she was intoxicated during a press conference. This summer, a PAC supporting Florida Governor DeSantis released an ad using AI-generated audio of former President Donald Trump attacking Iowa Governor Reynolds. While the audio was based on posts Trump had made on social media, it did not reflect words that were actually spoken by the former president. In a runoff election for president in Argentina this month, AI-generated images of candidates are circulating, both positive and negative, as well as manipulated videos of candidates making statements that were not uttered by the candidate. And on Jan. 23, 2024, voters in New Hampshire received phone calls that used AI generation technology to impersonate the voice of President Biden and urged them not to vote in that day’s primary election.