Can AI make political campaign ads fairer, or will it be used to spread lies?

Jonathan Rinaldi, a candidate for city council in Queens, New York, used his phone to ask an AI chatbot to create fake news stories and endorsements he had not actually received. He shared one of these stories, made to look like it was from CNN, on social media. It falsely claimed his opponent, Lynn Schulman, had quit the race.

Schulman had not quit and won the election in November. Rinaldi called his creations art and protected political speech. However, officials accused the 47-year-old of misleading voters. He was arrested on charges of forgery, which seems to be one of the first times someone has faced criminal charges for using AI in political advertising.

Rinaldi disagrees, stating he did nothing different from other campaigns. While such charges are rare, the case highlights the ongoing debate about how to regulate AI in political communication. He believes people should not be arrested for social media posts made by AI, citing free speech concerns.

AI-generated ads are now common in campaigns, often supporting candidates or criticising opponents, frequently paid for by outside groups. The widespread use of these tools, and the potential for misuse by some individuals, is raising concerns that false political content could influence voters and spread misinformation, especially before important elections.

AI-generated ads have already caused controversy in primary races. One ad showed a Texas candidate singing about trans kids dressed as a character from The Sound of Music. Another suggested a Kentucky congressman was in a relationship with other politicians. An LA mayoral candidate used AI to portray his opponent as the Joker, which helped him gain more attention.

Experts like Bruce Schneier suggest these ads are often about showing support for one's own group rather than convincing others. Even politicians like Donald Trump have used AI, creating images of himself as a king or a Jedi, and sharing questionable content. He has also shared AI-generated images of opponents and posted videos claiming to have cured critics of 'Trump derangement syndrome'.

Some AI uses have been more deceptive. For instance, Ron DeSantis's campaign shared an AI image that appeared to show Trump hugging Dr. Anthony Fauci. In New Hampshire, AI-generated robocalls mimicking Joe Biden's voice tried to discourage people from voting in the primary. These sophisticated tools make it much easier to create and spread false information quickly.

A recent poll indicated that a large majority of Americans believe AI-generated political content will spread misinformation during upcoming elections. This distrust is shared across different political parties.

Organizations like the Brennan Center for Justice are urging the Federal Election Commission to address deceptive AI content, fearing deepfakes could be used to trick donors. While US law prohibits fraudulent misrepresentation in political campaigns, enforcement can be difficult. Many states have introduced laws regulating deepfakes in political ads, often requiring disclosures or banning their use close to elections. However, some bans have been challenged in court for violating free speech rights.

Nathan Sanders, a data scientist, notes that the rules for political communication are already confusing, leading to inconsistent enforcement.

Rinaldi's campaign allegedly used AI to create images of his opponent and falsely claim endorsements from groups like a police precinct and a Jewish organization. When confronted about a fake endorsement image, Rinaldi dismissed it as 'politics'. Experts acknowledge that lying in politics is not new; Benjamin Franklin published false news for political reasons centuries ago. However, AI makes lying at a massive scale much easier, though the core issue remains the act of lying itself.

While some consultants acknowledge the risks, they also see AI as a democratizing force. It allows candidates with smaller budgets to create high-quality content similar to that of wealthier opponents. Julie Sweet of the American Association of Political Consultants (AAPC) states AI can be used for both truthful and untruthful content. The AAPC recommends disclosing the use of AI in political ads, especially when it creates realistic fake audio or video, but suggests simpler uses like translation may not require disclosure.

Some believe that mandatory disclaimers like 'fake' or 'manipulated' could reduce audience trust even for non-deceptive content. Instead, they suggest terms like 'dramatization' or 'synthetic voice'.

A tool called BattlegroundAI is designed to help users create political ads that comply with current regulations, including the duration of disclaimers. Its founder believes that while AI might lower the barrier for bad actors to lie, it doesn't represent everyone using the technology.

Researchers anticipate a significant increase in AI-generated political content. With easier access to free AI tools, candidates can now produce personalized campaign materials quickly, as Rinaldi did from his home. He is now running for state assembly, aiming to use these tools against what he calls the 'power structure'.