Menu
SUBSCRIBE
SUBSCRIBE
Menu

Synthetic Media

Ad images, a shield, and a checklist beside a capitol in a keystone frame, for Pennsylvania's Deceptive Advertising Act

Pennsylvania Passes Bill Requiring Ads to Disclose AI Content That Could Fool Buyers

, , , , , , , , , ,

Ads with AI-made images, video, or audio that could mislead buyers would need a label saying AI was used, with fines of up to $3,000 for leaving it off.

 

Pennsylvania lawmakers passed a bill requiring advertisers to label ads that use AI-made or AI-altered images, video, or audio in a way that could mislead buyers.

The state House approved Senate Bill 806, the Artificial Intelligence Deceptive Advertising Act, by a vote of 163 to 40. The Senate passed the same text 48 to 2 in June, so the bill now goes to Gov. Josh Shapiro. He can sign or veto it. If he does neither within 10 days of receiving it, it becomes law without his signature. It would take effect 60 days after it becomes law.

Companies advertising in Pennsylvania would have to add a clear label to any ad where AI content could give a false impression of the product. If they don’t, the state attorney general or a county district attorney can take them to court, where a judge can impose a fine of up to $3,000 plus legal costs.

Pennsylvania joins New York and California, but goes further

New York began requiring labels on ads that show AI-generated people on June 9. The law carries a $1,000 fine for a first violation and $5,000 for each subsequent violation.

California followed on Sept. 16, when Gov. Gavin Newsom signed a law requiring a label on video and audio ads that use AI-generated performers.

Pennsylvania’s bill covers more kinds of content than either law. It applies to any image, video, or audio made or heavily changed by AI, not only AI-generated people. But it requires a label only when the content could mislead buyers about what is being sold, where the product comes from, or whether it is real.

Ads must say they contain AI content

The label must say that the ad contains content made or changed using AI. It must appear as soon as the AI content does, in the same form as that content: on-screen for video, spoken for audio, and printed for print ads. It must be easy to notice and understand, and in the language the ad uses. Where the technology allows, it must be permanent or very hard to remove.

The rule applies to advertisers that run such an ad knowingly or recklessly. Ads that are obviously fiction are exempt, as are parody, satire, commentary, and criticism. Ads made up only of text are also exempt.

The state can go to court, with fines up to $3,000

The attorney general or a district attorney can ask a court to order an advertiser to stop running an ad, and can sue over a violation. A court can fine the advertiser up to $3,000 and order it to pay legal fees and court costs.

Adding a label does not shield an advertiser from lawsuits under other laws.

Senators from both parties backed the bill

State Sen. Nick Pisciottano (D) sponsored the bill with Democratic and Republican senators. In a memo asking colleagues to join him, he wrote that AI images and video have become so realistic that the average shopper struggles to tell them apart from real ones, and that a label would help fight scams.

Shapiro has not yet signed the bill.

Nighttime view of Earth from space accompanying coverage of Google’s withdrawal of its Google Earth AI tool.

Google Pulls Google Earth AI Tool One Day After Launch

, , , , , , , , ,

The feature generated fake explosions, military sites, and conflict scenes before Google withdrew it to add stronger safeguards.

 

Google pulled its Google Earth AI tool one day after launch after people shared screenshots of generated images that appeared to violate its policies.

The company launched the “create image” feature globally for Google Earth web users on July 30. It combined Google Earth’s satellite, aerial and 3D imagery with the Nano Banana 2 image generator.

Users could select a real location, describe the scene they wanted and generate a new image based on that place.

Fake satellite images could be mistaken for evidence

The tool could create convincing images of events that never happened at real locations.

Agence France-Presse reported that its tests produced images of an explosion in Paris, a nuclear site in Iran, a bomb crater in Russia and an Islamic State training ground in Syria.

The tool also generated fabricated floods in Bangladesh, a migrant camp at the U.S.-Mexico border and a warehouse holding vans in Georgia.

Researchers warned that images like these could circulate online as supposed evidence of military activity, natural disasters or other real-world events. Google Earth imagery is commonly used by journalists and researchers to check images and videos tied to specific locations.

Initial controls did not prevent misuse

Google said the generated images were visible only to the people who created them. They did not replace satellite imagery in the main Google Earth service or appear to other users.

The images also included watermarks identifying them as AI-generated. Google said on X that it used SynthID, an invisible marker embedded in media created with its AI tools.

Those controls did not prevent people from taking screenshots and sharing them outside Google Earth. Once separated from the original interface, the images could be presented without the surrounding information showing how they were made.

Google said it withdrew the feature on July 31 because people “uniquely trust Google Earth for a reliable view of the world.” The company said it would add stronger safeguards but did not specify what those controls would include.

Google promoted planning and educational uses

Google originally promoted the feature as a way to recreate historic locations, visualize real estate developments and preview construction projects.

Examples included showing how Pompeii might have looked before the eruption of Mount Vesuvius and placing a proposed building on an empty lot.

Google has not said when the feature will return. The “create image” button remained unavailable after the company announced the rollback.