Disclosure on ads is a good first step—but it’s not enough.
-
At least five states have already moved to make campaigns tell voters when they use “AI” so we can decide how well we can believe our eyes. A few advertisements have already been rigged with fake voices and pictures to make viewers like or hate a candidate.
There’s a scene in the 1966 film of Ray Bradbury’s ominous novel “Fahrenheit 451” when main character Montag, flees from police in a dystopian dictatorship of the future and hides at a safe house, where he sees himself captured and killed on TV as a robotic voice assures viewers that everyone is safe again.
That was imaginative science fiction a half-century ago. But now it’s easy to create things that never happen, through the wonders of artificial intelligence. This isn’t the “fake news” Donald Trump rails against, but deliberately deceptive images and impressions in advertising meant to spread real whoppers to millions via social media and television.
That’s why at least five states have already moved to make campaigns tell voters when they use “AI” so we can decide how well we can believe our eyes. A few advertisements have already been rigged with fake voices and pictures to make viewers like or hate a candidate.
We saw the benign use of AI last year when Paul McCartney and Ringo Starr recorded a new song featuring two deceased Beatles. The New York Times even created a new editor position to use the technology fairly and efficiently. But AI has also given politicians new ways to lie in their advertising by manipulating voices and images.
It would be great if we could just outlaw it in political advertising, but that poses First Amendment problems. To encourage free-wheeling debate, the U.S. Supreme Court allows wide latitude for political speech, so it’s fair game to take one tiny detail and blow it out of proportion in a campaign. If I, or some supposedly “independent” campaign committee supporting me, want to post ads saying you favor gun control and want dirty books in school libraries, there’s not much you can do—except to find something nastier to say about me.
But what if I don’t just falsely twist what you said, but distort some documents and videos to spread some smears about your voting record or personal habits? Your rebuttal will never quite catch up with my lie and, even if it does, our race will probably be over by then.
In a well-intended first step toward curbing this bold new political pollution, Florida Sen. Nick DiCeglie has filed a bill requiring a disclaimer on advertising made with “generative artificial intelligence.” Rep. Alex Rizo, R-Hialeah, has a House companion bill.
“The increasing access to sophisticated AI-generated content threatens the integrity of elections by facilitating the dissemination of misleading or completely fabricated information that appears more realistic than ever,” DiCeglie, a Republican from Indian Rocks Beach, said of his legislation. “The technology that produces this content has advanced rapidly and outpaces government regulation.”
AI-contaminated advertising would have to be labeled, “Created in whole or in part with the use of generative artificial intelligence (AI).”
That’s something, and maybe all, the First Amendment would permit in court, but a consumer alert is not sufficient safeguard for the sewer that modern campaigning has become. An advertisement shouldn’t just say “contains AI,” it should specify which allegations are vicious fantasy.
“The part of this ad showing my opponent taking his children to the movies is accurate. The part that shows him laughing and applauding when the hunter kills Bambi’s mother is AI,” might be a suitably specific disclaimer.
Regrettably, any restrictions imposed nationally by Congress or state legislators won’t be so scrupulous. Voters need a bit of caveat emptor at the polls.
The best defense against fake advertising is to simply not believe anything one side says about the other— conservative or liberal, Democrat or Republican, veteran incumbent or upstart newcomer. Just remember, somebody is spending a lot of money to persuade you this way or that way about a person or an issue on the ballot, so everything in every advertisement is meant to stir emotions, not educate. And of course if anything seems too good— or too bad—to be true, it probably isn’t.
Bill Cotterell is a retired capitol reporter for United Press International and the Tallahassee Democrat. He can be reached at bcotterell@govexec.com.
