Pennsylvania enacts law on labeling deceptive AI ads

By VK

Published 25 Oct 2026, 11:44 UTC

Reporting from The Verge

Pennsylvania will require a clear AI label on ads that use generated or altered media in a misleading way, with civil fines of up to $3,000.

What it means for founders

Creative tool makers inherit the problem. The posted text puts the duty on advertisers rather than model vendors, but if you sell AI image, video or voice generation for marketing, customers running ads in Pennsylvania are likely to want a built-in way to add a compliant label. One-click disclosures, burned-in overlays and audio tags become product features, a pattern already visible when OpenAI added EU text watermarks.
Durability is the hard part. A caption layer anyone can delete may fall short of a label that is permanent or very hard to remove. Provenance metadata and overlays that survive cropping and re-encoding are worth a look.
Small advertisers are in scope. The posted text sets no size threshold, so local shops and small agencies on self-serve ad tools appear to be covered alongside national brands.
Watch late December and the first case. The first action by state or county prosecutors should show how broadly the deception test is read. Ad platforms that generate creative on the fly, like the visual ads OpenAI is testing in ChatGPT, are worth watching for how they label output.

The story

Governor Josh Shapiro put his signature on the Artificial Intelligence Deceptive Advertising Act on Thursday, October 22, The Verge reports. The measure, Senate Bill 806, was introduced by State Senator Nick Pisciottano. The details below come from the latest printed version the General Assembly has posted, amended June 10, 2026, which may not be the exact text Shapiro signed.

What the posted text requires

Under it, an advertiser who knowingly, recklessly or willfully makes and runs an ad, or has one made and run, breaks the rule if the ad's AI material misleads consumers about what is being sold, where it comes from or whether it is genuine, and carries no clear and conspicuous disclosure. Synthetic content means information such as images, video, audio or text that AI has generated or significantly modified.

In that case the ad needs a clear and conspicuous notice telling the audience that AI was used to create or alter some of its content. The notice must appear at the outset whenever the AI material is presented, use the same medium as that material and, where technically feasible, be permanent or very hard to strip out. Spoken notices need audible volume and cadence, on-screen ones must stay up long enough to read, and print ones must contrast with the background.

The posted text exempts three kinds of content: anything evidently fictional from its content or context; parody, satire, commentary or criticism; and content limited to text. That last carve-out contradicts The Verge, which lists text among the formats that need a label.

Enforcement and timing

The state Attorney General or the relevant district attorney can seek an injunction and bring a civil action. A court may impose a fine of up to $3,000 and award attorney fees and court costs. The posted text says the act takes effect in 60 days, which points to late December. The Verge says the rule applies wherever an ad was made, so national ads that also air in Pennsylvania could show labels in other states.

What we don't know yet

How enforcers will judge a significant AI edit, an evidently fictional ad or a false consumer impression, none of which the posted text defines.
Whether the $3,000 cap applies per ad, per airing or per case; the posted text is silent.
Whether the signed law matches the June 10 text the legislature has posted, including the text-only exemption.