Illustration: Nick Barclay / The Verge
Meta announced Wednesday that it would require advertisers to disclose when potentially misleading AI-generated or altered content is featured in political, electoral, or social issue ads.
The new rule applies to advertisements on Facebook and Instagram that contain “realistic” images, videos, or audio falsely showing someone doing something they never did or imagining a real event playing out differently than it did in reality. Content depicting realistic-looking fake people or events would also need to be disclosed. The policy is expected to go into effect next year.
“In the New Year, advertisers who run ads about social issues, elections & politics with Meta will have to disclose if image or sound has been created or altered digitally, including with AI, to show real people doing or saying things they haven’t done or said,” Nick Clegg, Meta president of global affairs, said in a Threads post Wednesday.
Content that’s been edited in ways “that are inconsequential or immaterial to the claim, assertion, or issue raised in the ad,” like cropping or color correcting, does not need to be disclosed, according to Meta’s Wednesday blog post.
For ads containing digitally altered content, Meta says it will flag the information to users and log it in Meta’s ads database.
Earlier this week, Reuters reported that Meta was banning political campaigns and groups from using its new slate of generative AI advertising products. The tools allow advertisers to create multiple versions of ads, including different backgrounds, text, and image and video sizing.
The decision to disclose AI-generated content in political ads comes as lawmakers and regulators are preparing to take on the issue themselves ahead of the 2024 presidential election.
Earlier this year, Rep. Yvette Clarke (D-NY) and Sen. Amy Klobuchar (D-MN) introduced bills requiring campaigns to disclose when ads include AI-generated content. The Federal Election Commission, the regulatory agency in charge of political advertising, is also expected to make a decision on a new rule requiring political campaigns to disclose the use of AI-generated content. It’s unclear exactly when a vote on this rule could take place.
