Google and YouTube allow political ads to use AI to show fake events

Google is allowing the use of artificial intelligence in political advertising, as long as advertisers admit to using it.

The Bay Area tech giant, which also includes YouTube and Waymo under its corporate umbrella, announced Wednesday that it will require election advertisers to include clear disclosures when their ads include “synthetic content that inauthentically depicts real or realistic-looking people or events,” according to new advertising policy language shared with SFGATE. The requirement goes into effect in mid-November, the company said, about a year before the 2024 presidential election. 

Google’s dominant search engine and video player see huge amounts of political spending, much like Facebook. Joe Biden’s presidential campaign has poured more than $83 million since May 2018 into Google and YouTube ads, according to OpenSecrets. Candidates and political action committees can target Google and YouTube audiences by geographic location, age and gender using Google’s audience targeting function.

The Mountain View-based tech giant is framing the policy as a “transparency” update meant to improve voters’ confidence in election ads. A Google spokesperson told SFGATE that the disclosures would be required for ads that make it appear as if a person is saying something they didn’t say, and for ads that show a realistic-looking but fake event. Advertisers who use digitally generated content will have to “prominently” include statements like, “This audio was computer generated,” or, “This image does not depict real events,” the spokesperson said. 

Google is both an advertising arbiter and an AI powerhouse, and the new disclosure policy exists in lieu of a ban on artificially created political content altogether. The company’s ad policies already included a rule against “deceptively doctoring media related to politics, social issues, or matters of public concern.” SFGATE asked whether that rule would change, and Google said it would not.

“We will continue to enforce all of our policies, including our manipulated media and election misinformation policies, wherever we find violations,” the representative said.

The company seems set to allow content it judges to be “inauthentic” but not “deceptive.” The spokesperson said the company may suspend advertiser accounts for “pervasive or egregious” violations of the new rule.

The rule change comes as enthusiasm about artificial intelligence continues to bloom across the tech world; it’s become easier than ever to quickly create synthetic audio, images or video. Chatbots like OpenAI’s ChatGPT, released last November, led the “generative AI” charge, and corporate giants have begun releasing new AI-based business products. Websites that offer to create fake images and audio clips have propagated across the internet. What may have once taken a skilled artist in Photoshop hours now requires a few seconds and a well-worded prompt.

Text-to-image startups use massive stores of data to imitate artistic or photographic styles and deliver whatever image a user asks for; text-to-audio tech can replicate the sound of people’s voices using just a few minutes of recorded speech. With a few keystrokes, anyone could make a computerized voice that sounds like Joe Biden’s say, “You should vote for Donald Trump.”

Already, Florida Gov. Ron DeSantis’ campaign used AI to impersonate Trump’s voice in a July campaign ad, according to Politico. A “Beat Biden” GOP ad in April used AI-generated images — some of which appeared to mimic San Francisco — to portray a country in disrepair. 

Hear of anything happening at Google or another tech company? Contact tech reporter Stephen Council securely at [email protected] or on Signal at 628-204-5452.