Members of the U.S. Senate are sounding the alarm about the threat that artificial intelligence poses to elections through its ability to deceive voters. But the prospects for legislation that can meaningfully address the problem appear uncertain.

In a Wednesday hearing, the Senate Rules Committee advanced three bills designed to counter the AI threat. But the only one to receive support from Republicans on the panel would simply create voluntary guidelines for election officials. It stops well short of restricting the use of AI in elections or even requiring disclosure of its use — steps that a growing number of states have already taken.

Sen. Amy Klobuchar, D-Minn., who chairs the Rules Committee and introduced all three measures, said the ability of generative AI to create deceptive images fundamentally threatens fair elections.

“We are going to see this resurgence of fakery and scams going on in our elections,” said Klobuchar. “And whether you’re a Democrat or a Republican, whether you’re a conservative or a liberal, we cannot have our democracy undermined by ads and videos where you literally don’t know if it’s the candidate you love or the candidate you dislike.”

Senate Majority Leader Chuck Schumer framed the stakes as even higher.

“Our democracy may never recover if we lose the ability to distinguish at all between what’s true and what’s false, as AI threatens to do,” said the New York Democrat, whose appearance at the hearing was a potential signal that he aims to prioritize the legislation.

AI robocalls, images

The dangers posed by AI were starkly illustrated in February, when thousands of New Hampshire voters received a robocall with an AI-generated voice impersonating President Joe Biden, urging them not to vote in the upcoming state primary. A Democratic operative working for a rival candidate has admitted to commissioning the calls.

And last June, the presidential campaign of Florida Gov. Ron DeSantis released a video that appeared to use AI-generated images of former President Donald Trump hugging Dr. Anthony Fauci, the former chief medical adviser to Biden who is deeply unpopular among GOP primary voters.

Klobuchar added at the hearing that she wanted to keep the issue out of the “partisan milieu.” Blue, purple and red states have lately passed laws to address the AI threat, she noted. And all three pieces of legislation Klobuchar has introduced have Republican co-sponsors.

Still, there were signs that avoiding partisan politics could prove impossible.

Republicans in Congress, often led by House Speaker Mike Johnson of Louisiana, have strongly opposed previous efforts by the Biden administration to restrict the spread of political disinformation, saying they violate speech rights and give too much power to government regulators.

Sen. Deb Fischer, R-Neb., the ranking Republican on the panel, raised similar concerns Wednesday about two of the Klobuchar bills: the Protect Elections from Deceptive AI Act, and the AI Transparency in Elections Act.

The Protect Elections from Deceptive AI Act would bar deceptive AI video or audio relating to candidates for federal office. It was introduced in September by Klobuchar and has five co-sponsors, including Sen. Josh Hawley, R-Mo.

The AI Transparency in Elections Act requires that political ads that use AI contain a statement disclosing its use. That measure, introduced by Klobuchar in March, is co-sponsored by Sen. Susan Collins, R-Maine.

The two bills, Fischer argued, “increase burdens on speech,” and are too vague in defining AI, creating uncertainty about whether a speaker might be subject to penalties. They also aim to federalize the issue and pre-empt state law, encroaching on state control of elections, Fischer added.

Both measures were passed out of the Rules Committee Wednesday on party-line votes, with no Republican support. (The official vote count in both cases was 9-2, because some Republicans who didn’t attend the hearing voted ’no by proxy,’ which isn’t counted as an official vote.)

The third measure, the Preparing Election Administrators for AI Act, was passed out of committee unanimously.

It would require the U.S. Election Assistance Commission to consult with the National Institute of Standards and Technology in creating voluntary guidelines for election officials for how to protect against the threat of AI in elections, especially with respect to its use by foreign adversaries.

On May 13, Reps. Brian Fitzpatrick, R-Penn., and Chrissy Houlahan, D-Penn., introduced companion legislation in the House.

Over the last year or more, states including Texas, Florida, New York, Michigan, Indiana, Wisconsin, Oregon, Utah, New Mexico and Idaho all have passed legislation restricting the use of AI in political ads or requiring that it be disclosed.