You’re reading Superintelligent, a newsletter about AI. Click here to get the full edition in your inbox.
Darrell M. West is a senior fellow at the Brookings Institution’s Center for Technology Innovation and co-editor in chief of TechTank.
Texas voters might have come across a video last month of Democratic Senate candidate James Talarico reading controversial tweets that he posted over the past decade. “So true,” he said of his own comments. “I love this one, too,” he said at another point.
The video clearly was not real. It was an ad created by the National Republican Senatorial Committee meant to highlight “extreme statements praising transgenderism, twisting Christian beliefs, and advocating for open borders.” The rendition includes a label noting that it was generated with artificial intelligence and begins with a narrator describing the video as a “dramatic reading” of the candidate’s comments.
But voters should be aware of the coming onslaught of similar ads that will likely go much further in falsifying candidate statements, manufacturing persuasive fake videos and putting campaigners in embarrassing situations that did not actually take place. Deepfakes have advanced to the point where even experts cannot easily discern true from false. The prior deficiencies of AI-generated content, such as hands with extra fingers or fake heads put on actual bodies, have given way to sleek videos and pictures that are completely believable.
What are voters to do in this new world? Labeling ads as AI-generated is a popular remedy based on the idea that telling people something is AI-generated will reduce the seriousness of the offense. Just as legislators mandated candidates to add language to campaign spots informing voters “I paid for this message” to improve transparency and accountability, companies and some states are starting to require ad-makers to inform viewers when their advertisements rely upon AI.
Though helpful from a voter standpoint, labeling does not deal with the problem of ads conveying fictitious or erroneous information. Candidates may suffer damage to their personal reputations in ways that hurt their electoral prospects. Telling voters an ad is AI-generated does not negate that harm.
For this reason, Minnesota legislators have enacted rules to punish anyone who “falsely depicts a person without their consent with the intent to damage a political candidate’s reputation or skew election outcomes.” In the same way that false or misleading commercial ads can lead to product liability claims, reformers want to give candidates the option of holding opponents accountable for deepfakes that harm their political reputations.
Other places are prohibiting the use of deepfakes close to an election. For example, in Texas, candidates cannot run ads with deepfake videos in the 30 days prior to an election. That rule is designed to limit the harmful effects on voters right before they cast their ballots.
However, it is not clear whether these or related proposals will pass constitutional muster. Judges have long ruled that campaign speech is free speech and is therefore protected even if statements made in ads or on the campaign trail are false.
This leaves voters with few options to protect themselves from the increasing sophistication of generative AI tools and video editing software, which candidates will inevitably use to push the envelope way beyond the Talarico ad. In the coming months, voters will almost certainly see an onslaught of fake videos, manufactured images and misleading claims that make it difficult to discern the truth.
What voters need is digital literacy training that can help them spot digital fraud and abuse in many different arenas. Such training is being undertaken by a range of educational and nonprofit organizations, and it helps educate people about the way videos can be manipulated, how to evaluate sources of information, the importance of relying upon multiple sources of information and how to build personal safeguards against partisan messaging.
Fact-checkers such as PolitiFact and FactCheck.org can be helpful in evaluating political claims. They are nonpartisan in nature and regularly assess falsehoods from candidates and politicians. Many reporters and nonprofit organizations rely upon their conclusions regarding factual accuracy.
Tech companies also have responsibilities to help voters. They should take content moderation seriously, use AI to spot deepfakes, employ human checkers to identify blatantly false claims, refuse to amplify fake messages and take down content that violates their terms of service.
Voters should be especially careful about campaign messaging or social media posts emanating from abroad. Foreign influence operations have organized disinformation campaigns targeting American elections and spread falsehoods about a range of different topics. Their goal is to sow public mistrust and create uncertainty about what is true.
As the 2026 campaign season unfolds, voters will need a “buyer beware” mentality that embraces skepticism and disbelief. If some comments seem fishy or images appear misleading, people should verify the information before passing it along to others. The upcoming election will test voter abilities to evaluate information and discern false claims.
The post Can you spot a fake political ad? AI is making it harder. appeared first on Washington Post.
