Days after New Hampshire voters received a robocall with an artificially generated voice similar to that of President Joe Biden, the Federal Communications Commission has banned the use of artificial intelligence-generated voices in robocalls.

This is a breaking point. The 2024 U.S. election will be the first to be held with widespread public access to AI generators that allow people to create images, audio, and videos—some of them for nefarious purposes.

Agencies are rushing to limit misconduct caused by artificial intelligence.

Sixteen states have enacted legislation targeting the use of artificial intelligence in elections and campaigns; many of these states require disclaimers on synthetic media close to elections. The Election Assistance Commission, a federal agency that supports election administrators, has released an “artificial intelligence toolkit” with tips election officials can use to convey election messages in an age of fabricated information. States have launched their own pages to help voters identify AI-generated content.

expert warn About the potential for AI to create deepfakes that make candidates appear to say or do things they didn’t. Experts say AI’s influence could harm the United States at home — misleading voters, influencing their decisions or preventing them from voting — and abroad, benefiting foreign adversaries.

But the anticipated avalanche of AI-driven misinformation never materialized. As Election Day came and went, viral misinformation played a starring role, misleading information about vote counts, mail-in ballots and voting machines. However, this deception relies heavily on old, familiar techniques, including text-based social media statements and videos or images taken out of context.

“It turns out that using generative AI is not necessary to mislead voters,” said Paul Barrett, associate director of NYU’s Stern Center for Business and Human Rights. “This is not an ‘AI election.'”

Daniel Schiff, an assistant professor of technology policy at Purdue University, said there wasn’t any “massive last-minute campaign” to mislead voters at polling locations and hurt turnout. “This misinformation is smaller in scope and at least less likely to be a decisive factor in the presidential election,” he said.

Experts say the AI-generated claims that get the most attention support existing narratives, rather than fabricating new ones to fool people. For example, former President Donald Trump and his vice presidential running mate J.D. Vance falsely claimed Haitians eat pets in Springfieldohio, artificial intelligence images Memes depicting animal cruelty flood the internet.

Technology and public policy experts, meanwhile, say safeguards and legislation minimize the potential for AI to generate harmful political speech.

Schiff said the potential harm of artificial intelligence to elections has spurred “urgent energy” focused on finding solutions.

“I believe a high level of concern from public advocates, government actors, researchers and the general public is really important,” Schiff said.

YuanThe company that owns Facebook, Instagram and Threads requires advertisers to disclose the use of artificial intelligence in any ads about political or social issues. TikTok applies a mechanism to automatically tag some AI-generated content. OpenAI, the company behind ChatGPT and DALL-E, banned the use of its services for political campaigns and blocked users from generating images of real people.

Misinformation actors use traditional techniques

Siwei Lu, a professor of computer science and engineering at the University at Buffalo and an expert in digital media forensics, said the power of artificial intelligence to influence elections is also diminishing because there are other ways to gain such influence.

Herbert Chang, an assistant professor of quantitative social science at Dartmouth College, said: “The impact of artificial intelligence may appear weak in this election because traditional forms are still more effective and social-based platforms such as Instagram On online platforms, accounts with a large number of fans are less likely to use artificial intelligence.” Chang co-authored a study that found that AI-generated images “generate less virality than traditional memes,” but memes created with AI can also generate virality.

Celebrities with large followings can easily spread their message without AI-generated media. For example, Trump has repeatedly falsely stated in speeches, media interviews, and on social media that illegal immigrants are being persecuted. Bring to America to vote Although it is extremely rare for non-citizens to vote, and citizenship is required to vote in federal elections. Polls show Trump’s repeated claims are paying off: More than half of Americans in October said they were concerned about non-citizens voting in the 2024 election.

PolitiFact, which fact-checks and stories about election-related misinformation, points to some images and videos that use artificial intelligence, but much of the viral media is what experts call “cheap fakes” — deceptions that have not been edited by AI. Sexually authentic content.

In other cases, politicians are flipping the script — blaming or belittling AI rather than using it. For example, Trump falsely claimed that a montage of his gaffe posted by the Lincoln Project was AI-generated, and he also said that a group of Harris’ supporters were AI-generated. CNN reported that North Carolina’s Lieutenant Governor Mark Robinson After posting offensive comments on a porn forum, Robinson claimed it was artificial intelligence. One expert told WFMY-TV in Greensboro, North Carolina, that the scenario Robinson claimed was “almost impossible.”

Artificial intelligence once fomented ‘partisan hostility’

Authorities discovered that a street magician in New Orleans created fake Biden robocalls in January in which the president could be heard discouraging people from voting in the New Hampshire primary. The magician said it only took him 20 minutes and $1 to create the fake audio.

A political consultant who hired a magician to make phone calls faces a $6 million fine and 13 felony charges.

It’s a striking moment, in part because it’s not repeated.

In the weeks leading up to Election Day, AI was not driving the spread of two major misinformation categories—fabricated pet-eating claims and lie Bruce Schnell, an adjunct lecturer in public policy at Harvard Kennedy School, said about the Federal Emergency Management Agency’s relief efforts after hurricanes Milton and Helen.

“We’ve certainly witnessed the use of deepfake technology that appears to be very effective in stoking partisan hostility and helping to establish or solidify certain misleading or false perceptions about candidates,” said Daniel Schiff.

He teamed up with Kaylyn Schiff, an assistant professor of political science at Purdue University, and Christina Walker, a Purdue doctoral student, to create a database of political deepfakes.

Data shows that most deepfakes are intended to be satirical. Behind it are deepfakes designed to damage someone’s reputation. The third most common deepfakes are created for entertainment.

Daniel Schiff said deepfakes that criticize or mislead people about candidates are “an extension of traditional American political narratives,” such as portraying Harris as a communist or a clown or Trump as a fascist or criminal. Zhang agreed with Daniel Schiff, saying that generative AI “exacerbates existing political divisions, not necessarily to mislead, but through exaggeration.”

Major foreign influence operations rely on actors, not AI

Researchers warned in 2023 that artificial intelligence could help foreign adversaries conduct influence operations faster and cheaper. The Center for Foreign Malignant Influence, which assesses foreign influence activities against the United States, said in late September that AI had not “revolutionized” those efforts.

To threaten U.S. elections, foreign actors must overcome the limitations of artificial intelligence tools, evade detection and “strategically target and spread such content,” the center said.

Intelligence agencies — including the Office of the Director of National Intelligence, the FBI, and the Cybersecurity and Infrastructure Security Agency — flag foreign influence operations, but they more often employ actors in staged videos. A video shows a woman claiming Harris hit her in a hit-and-run crash. The video’s narrative is “completely fabricated,” but not artificial intelligence. Analysts linked the video to a Russian network called Storm-1516, which used similar tactics in the video to try to undermine election trust in Pennsylvania and Georgia.

Platform safeguards and national legislation may help curb ‘the worst behaviour’

Social media and artificial intelligence platforms are trying to make it harder to use their tools to spread harmful political content by adding watermarks, labels and fact-checks to claims.

Meta AI and OpenAI both said their tools rejected hundreds of thousands of requests to generate AI images of Trump, Biden, Harris, Vance and Democratic vice presidential candidate Minnesota Gov. Tim Walz . In a report on the 2024 global elections released on December 3, Nick Clegg, Meta’s president of global affairs, said: “AI content related to elections, politics and social topics is rated only 1%. Less than 1% of all misinformation is fact-checked.”

Still, there are some drawbacks.

The Washington Post found that ChatGPT still wrote campaign messages targeted at specific voters when prompted. PolitiFact also found that Meta AI easily generated images that could support the claim that Haitians eat their pets.

Daniel Schiff said that as artificial intelligence technology advances, these platforms still have a long way to go. But at least through 2024, their precautionary measures and state legislative efforts appear to be paying off.

“I think strategies like deepfake detection, public awareness, and outright bans are all important,” Schiff said.