by Maria Wood
In January 2024, in the midst of the Democratic presidential primary season, some voters in New Hampshire received a robocall supposedly from President Joe Biden urging them to save their vote for the November election.
Here’s what the call said: “This coming Tuesday is the New Hampshire Presidential Preference Primary. Republicans have been trying to push nonpartisan and Democratic voters to participate in their primary. What a bunch of malarkey. We know the value of voting Democratic when our votes count. It’s important that you save your vote for the November election. We’ll need your help in electing Democrats up and down the ticket. Voting this Tuesday only enables the Republicans in their quest to elect Donald Trump again. Your vote makes a difference in November, not this Tuesday.”
While the robocall may have appeared to be a typical campaign communication, the voice on the call was not President Biden and the information was not true. The creators of the call used artificial intelligence, or AI, to create his voice and spread misinformation. Artificial intelligence is essentially computer technology that has the ability to mimic human reasoning and, as this call indicated, human voices. AI can also produce high quality fake images as well.
The League of Women Voters of New Hampshire filed a lawsuit alleging the robocall violated federal law, as well as a number of state laws. The suit sought to impose a fine and stop the creators of the robocall from producing similar media.
The fake robocall also caught the attention of the Federal Communications Commission (FCC). In February 2024, the FCC issued a declaratory ruling that made AI-generated voices used in robocalls illegal.
“Bad actors are using AI-generated voices in unsolicited robocalls to extort vulnerable family members, imitate celebrities, and misinform voters,” FCC Chairwoman Jessica Rosenworcel said in a statement. “State Attorneys General will now have new tools to crack down on these scams and ensure the public is protected from fraud and misinformation.”
This robocall is just one example of what are called deepfakes—audio and/or visual media made with AI tools that can simulate photos from events that never happened or create audio statements that were never said by a candidate. The potential danger is that voters may be swayed by false information about a candidate or dissuaded from voting altogether.
Ability to deceive
The capabilities of AI have grown in recent years, reaching a level of sophistication that could only have been dreamed of just a few years ago. David Opderbeck, a professor at Seton Hall Law School and co-director of its Institute for Privacy Protection, suggests that sophisticated tools that are readily available now make it easier to generate AI content that seems realistic, swamping already overloaded voters.
“Today, AI allows the substance of the content to be generated and manipulated in ways not possible before. One of the biggest dangers is that it makes it easy to generate content that seems authentic and seems as if it’s coming from a reliable source but isn’t,” says Professor Opderbeck, who is also co-director of the Gibbons Institute for Law Science and Technology. “Unless you are really aware of what you are consuming it’s easy to be deceived. Combine that with the sheer scale and volume that automated AI produces, and that is what many are rightly concerned about.”
Ellen Goodman, a professor at Rutgers Law School in Camden, maintains AI is not necessarily a bad thing. Campaigns, she notes, frequently use AI to sharpen and micro-target their messages to certain voters. It is when AI is used to amplify misinformation that voters can be easily misled.
“If you are giving wrong information about when and how to vote, or whether an election can be trusted—that can be a danger,” says Professor Goodman, who is the co-director of the Rutgers Institute for Information Policy & Law.
States, federal government take action
While campaigning has always veered toward sometimes sharp personal and misleading attacks between candidates, AI has increased the speed with which those attacks and false information can spread over social media. In response to AI technology, several states and the U.S. Congress have proposed and/or enacted laws attempting to regulate AI in campaigns.
Michigan, for example, enacted a law prohibiting the use of AI-generated content in political advertising without disclosing the source of the material. Likewise, California has banned the distribution of what it terms deceptive media with the intent to harm a candidate or deceive voters within 60 days of an election.
While not specifically addressing AI in elections, the New Jersey Senate is currently considering two bills regarding deepfakes and AI. One would establish a Deepfake Technology Unit to “provide technical aid, expertise, education, and training concerning the investigation and analysis of potentially deceptive audio or visual media to state and local law enforcement agencies, prosecutors, and state courts.” That bill is currently being reviewed by a New Jersey Senate committee.
The other piece of New Jersey legislation would mandate criminal penalties for the dissemination of deceptive audio or visual media, or deepfakes. The bill was voted out of committee but has not had a hearing by the entire New Jersey Senate.
New Jersey State Senator Andrew Zwicker says he voted yes on the bill to establish penalties for spreading deepfakes because of the rapid rise in AI technology. Current technology makes it “difficult if not impossible for a viewer to know whether it is a genuine video or a manufactured one,” he says.
Meanwhile, the U.S. Senate is mulling a trio of bills attempting to regulate the use of AI in elections. The Protect Elections from Deceptive AI Act would prohibit “materially deceptive” AI-generated content in federal elections. The AI Transparency in Elections Act would mandate disclosure of AI-generated content in political advertisements. Lastly, the Preparing Election Administrators AI Act would require the Election Assistance Commission to work with the National Institute of Standards and Technology to report to Congress how AI is used in elections as well as create/establish guidelines for state and local election officials. At press time, all three bills were awaiting a vote on floor of the U.S. Senate.
AI and the First Amendment
Regulating AI in campaigns does have First Amendment implications. The concern is that regulations could infringe on free speech. Citing those concerns, Louisiana Governor Jeff Landry vetoed a bill in July 2024 that would have outlawed deceptive videos and images in political campaigns.
“While I applaud the efforts to prevent false political attacks, I believe this bill creates serious First Amendment concerns as it relates to emerging technologies,” Governor Landry wrote in a letter explaining his veto.
Professor Opderbeck agrees that regulating the actual content created by AI may clash with the First Amendment.
“The heart of the First Amendment is political speech and we do have to be careful about how we regulate in that area,” he says.
However, Professor Opderbeck notes that the First Amendment doesn’t protect against fraud. “So, I don’t think a law against deepfakes in elections necessarily violates the First Amendment if it’s getting things that are truly fraudulent.”
For instance, spreading false election information or interfering with voting is not protected by the First Amendment. In October 2023, Douglas Mackey was sentenced to seven months in prison following his conviction for conspiracy to interfere in the 2016 election. Mackey posted fake advertisements on his Twitter account telling Hillary Clinton supporters they could vote by text message.
Do social media companies have a role?
Election misinformation is a global problem. In addition to the U.S. election in November, national elections in more than 60 countries have taken place or will take place in 2024. Misinformation regarding these elections could potentially affect more than four billion voters globally.
As Senator Zwicker points out, voters today get much of their information online. Therefore, he says, social media companies should have a role in monitoring false or misleading content on their channels. In February 2024 at a security conference in Munich, 20 tech giants including Amazon, Google, LinkedIn, Meta (parent company of Facebook), Microsoft, OpenAI, Tik Tok and X (formerly Twitter), signed a voluntary accord pledging to monitor AI-created content as it pertains to elections.
In a statement, Brad Smith, vice chair and president of Microsoft said, “As society embraces the benefits of AI, we have a responsibility to help ensure these tools don’t become weaponized in elections.”
In August 2024, The Washington Post reported that X’s AI chatbot, called Grok, spread false information after President Joe Biden’s announcement that he was exiting the presidential race. The chatbot suggested Kamala Harris was not eligible for the 2024 presidential ballot, which is not true. An answer generated by Grok stated, “The ballot deadline has passed for several states for the 2024 election,” citing nine states in particular—Alabama, Indiana, Michigan, Minnesota, New Mexico, Ohio, Pennsylvania, Texas, and Washington.
The secretaries of state in four of those states—Michigan, Minnesota, New Mexico, and Pennsylvania—sent a letter to X owner Elon Musk, urging him to “immediately implement changes to X’s AI search assistant, Grok, to ensure voters have accurate information in this critical election year.”
Grok’s false narrative was corrected on July 31—10 days after President Biden’s announcement. While Grok is only available to Premium and Premium+ subscribers, the false information was “shared repeatedly in multiple posts—reaching millions of people,” according to the letter.
Senator Zwicker suggests the federal government mandate that social media companies label any altered media. One suggestion has been to mandate that AI-generated content carry watermarks to differentiate it from authentic content.
“Put the burden on them to keep up with the technology so that when the public is viewing something on social media, they are at least aware it is manipulated,” says Senator Zwicker.
Professor Goodman agrees more transparency is needed to help the public distinguish between AI-content and the real thing.
“If it’s AI-generated content in connection to electoral information, it should say so,” Professor Goodman says.
Discussion Questions
- Besides a robocall, give two examples of AI-generated content that could be harmful to a political campaign or to election integrity in general. Explain how the examples could cause harm.
- What do you think of the suggestion to label AI-generated content with watermarks? Will that solve the misinformation problem? Explain your answer.
Glossary Words
conviction— the result of a criminal trial where the accused or defendant is found guilty beyond a reasonable doubt.
deepfake—an image, video or audio recording that has been digitally altered.
nonpartisan— not adhering to any established political group or party.
veto — to refuse approval or passage of a bill that has been approved by a legislative body. The executive branch of government has the power to veto, but that power may be overridden with enough support.
This article originally appeared in the Special 2024 Election Edition of The Legal Eagle.