Pennsylvania wants ads to disclose AI use, but what exactly does that mean?

From Philly and the Pa. suburbs to South Jersey and Delaware, what would you like WHYY News to cover? Let us know!

In the virtual world of Instagram, Djour Philly cloaks his identity with a pseudonym — and a Mexican wrestling mask — to better report on Philadelphia’s food scene. Just don’t ask him to accept restaurant advertising that features AI-generated images.

“It becomes indistinguishable what’s real [and] what’s fake,” he said. “How can anyone decide they want to eat out when they’re getting sold a bill of false goods even from the initial look of the place?”

In posts on the topic, the influencer has called out the negative effects AI has on the local food scene’s creativity and originality. It’s also bad, he says, for consumer confidence.

Those objections are not unique. Over the past few months, multiple local restaurants have come under fire online after being accused of using AI images in their signage, social media posts and ads.

“I don’t get how any restaurant wouldn’t wanna show off their own product over AI garbage,” one Redditor wrote recently.

“Shame anyone who uses AI instead of hiring a local photographer,” commented another.

“Society has truly lost the plot. Using AI to show the food you make instead of actually taking pictures of said food!!!” a third person wrote.

Some of the accused restaurants have denied using AI — but starting Dec. 21, that may no longer be an option.

A new state law, the Artificial Intelligence Deceptive Advertising Act, signed by Gov. Josh Shapiro on Oct. 22, requires that AI use in ads be disclosed, but only under certain circumstances.

“It’s not every minor digital edit,” said the law’s sponsor, state Sen. Nick Pisciottano, a Democrat who represents parts of Allegheny County in Senate District 45 “It’s specific to AI-generated content that creates a false consumer perception.”

What exactly that means, however, and how enforcement will be applied, remain open questions.

The gap between the AI in advertising law’s intention and its provisions

Pisciottano’s interest in developing the law,one of the first of its kind in the country, came in response to seeing exponential growth in AI-generated content online.

“It was becoming more and more difficult to distinguish photos and videos that were real, created with traditional means, or created by AI,” Pisciottano said. “And I felt that that was a danger to consumers, and it would make it easier for people to get scammed or ripped off, and we should step in and make some kind of reasonable rules and regulation around AI to make sure folks know [whether] what they’re seeing is either generated by AI or not.”

In a Facebook post announcing his signing of the law, Shapiro articulated transparency as the main goal: “Pennsylvanians deserve to know if and when AI is being used in ads — and now, thanks to this new law, any advertisement in the Commonwealth that uses AI technology must make that crystal clear.”

Pisciottano made a similar statement in a news article announcing the law’s passage: “If you see an advertisement on Facebook or the internet somewhere and it was generated by AI, they have to put a disclaimer at the bottom that says this was generated by AI.”

But the law itself details a narrower application.

“What [the law] says and what its goal might be could be different things,” said Libby O’Neill, an advertising and AI attorney with the law firm Loeb & Loeb. “The sponsor has described this as, ‘If it’s AI, label it.’ The law, on the other hand, only requires a label if that AI use creates a false perception about the product, and we don’t really know what ‘false perception’ means in this context, or what use cases of AI actually are going to trigger a violation of this statute.”

That ambiguity may stem from the fact that the “false perception” language was added later, according to an earlier, amended copy of the bill’s text. In fact, the bill was originally conceived as a mandate that all AI content — not just content present in ads — include a disclosure. The law’s scope was later limited to advertisements and further narrowed to only cover AI content that creates a “false consumer perception.”

That standard may end up being redundant. There are already laws on the books, both statewide and federally, prohibiting advertising that misleads consumers or creates a false impression.

“And by the way, a label doesn’t fix that,” O’Neill said. “An AI label is not going to fix it if you’re being otherwise misleading. And so the question really becomes, what work is the ‘AI’ label doing? And if it’s really just to inform people that AI was used, then that kind of raises the question of why ‘false perception’ needs to come into play.”

So what, then, does the phrase “false consumer perception” actually mean? And in what circumstances would enforcement be triggered?

“I think that the short answer is we do have to wait to see how this plays out in court,” O’Neill said. “And it’s worth noting that some opposition speakers against the bill, when we look at the legislative history, pointed out this exact issue — that we don’t know false as compared to what.”