By This Hour Technology Desk
Pennsylvania will require certain advertisements made or substantially altered with artificial intelligence to carry disclosures, a new state law that reaches across still images, video, audio and text. The measure is intended to give consumers notice when synthetic material is used in advertising in a way that could shape their perception of a product or service.
The rule’s breadth is its most consequential feature. It is described as applying not only to polished television campaigns from large companies, but also to routine promotional material made by smaller businesses. Yet the same language that gives the law a wide potential reach also leaves major questions about when a disclosure is triggered, what the notice must communicate, and who will decide whether a particular ad crosses the line.
The Verge reported that Governor Josh Shapiro signed the Artificial Intelligence Deceptive Advertising Act on Thursday. The account describes Pennsylvania as among the earlier states to impose disclosure requirements specifically on AI material in advertising. It also suggests the statute could affect campaigns produced beyond the state’s borders if they are expected to run in Pennsylvania.
A disclosure rule that follows the medium
Under the reported terms, an advertisement that has been significantly modified or generated by artificial intelligence must include a label in the same format as the advertisement. That requirement matters because advertising increasingly moves among formats: a campaign may include written copy, a static graphic, a short video and an audio component, each tailored for a different channel.
A same-format rule appears designed to prevent a disclosure from being separated from the material it describes. For a visual ad, the notice would be visual; for audio, the disclosure would be audio; and for text advertising, it would be textual. The supplied account does not set out the label’s precise wording, length, placement, prominence or timing. Those details could determine whether consumers actually see or hear the disclosure before the advertisement’s main message has its intended effect.
The law’s reported coverage of still imagery is significant because a single image can communicate claims about a product’s appearance, performance, endorsement or setting. Video and audio add further possibilities: edited depictions can show a product in use or make it appear that a person has said something. Text, too, can be generated or extensively changed by AI systems. By naming all four forms, the statute appears to target advertising’s core modes rather than a narrow class of so-called deepfake media.
Still, the reported standard is not that every advertisement touched by an AI tool must be labeled. The threshold is material that has been “significantly” modified or generated by artificial intelligence, as summarized in the source material. That distinction places substantial weight on the meaning of significant modification. A law can set a broad principle, but advertisers need to know whether ordinary production steps using AI-assisted tools are covered or whether the requirement is focused on more consequential changes to content.
The central test is a consumer’s perception
The available account indicates that the measure is aimed at synthetic content used in a way that creates a false consumer perception about a product or service. That formulation narrows the apparent purpose of the law from AI use in general to deception in commercial messaging. It also makes the law’s application dependent on a judgment about what an audience would understand from an advertisement.
That judgment will not always be straightforward. An ad can be stylized, exaggerated or heavily edited without necessarily misleading its intended audience. Conversely, an artificial alteration may be subtle while still influencing how a consumer understands a product. The source material does not provide a detailed definition of a false consumer perception, nor does it explain the process by which disputed advertisements would be assessed.
These unanswered questions matter because the law’s obligations are linked to both the nature of the AI intervention and the likely meaning of the finished ad. A business may need to consider whether AI was used to create an asset, whether the result materially changes what is being represented, and whether the representation could lead consumers to a mistaken conclusion. The supplied information does not establish how those considerations will be weighed, or whether state guidance will clarify them.
There is a practical reason for the focus on consumer understanding. A disclosure requirement is not merely a technical obligation for marketing teams; it is a way of changing the information that accompanies a sales message. The apparent policy choice is that, in some circumstances, viewers or listeners should be told when an ad’s content is synthetic or significantly altered. But the law’s reported wording leaves room for disagreement about the boundary between a creative production technique and an alteration that demands notice.
Exemptions leave room for interpretation
The statute reportedly contains exceptions for parody, satire, commentary, criticism and material that is plainly fictional. Those carve-outs recognize that not every artificial or altered depiction is presented as literal commercial reality. They also create a second layer of interpretation alongside the question of whether a piece of advertising is significantly AI-modified.
“Evidently fictional” is especially consequential language because fictional presentation can be obvious to one audience and less obvious to another. The source account does not specify a test for deciding when fictional character is sufficiently clear, or identify how an ad that mixes ordinary product claims with fanciful AI-made imagery would be treated. The exceptions may protect a broad range of expressive material, but their precise limits are not clear from the available information.
The reported exemptions for commentary and criticism raise similar questions of classification. Advertising can use humor, cultural references and critical themes while retaining a commercial purpose. The information provided does not explain how the statute distinguishes those categories when they overlap, or whether the exception turns on the content itself, the advertiser’s purpose, the audience’s likely impression, or some combination of those factors.
Those uncertainties do not mean the law lacks force. Its stated requirement is clear at a high level: certain AI-generated or substantially AI-modified ads must disclose that fact. But they do mean that the practical reach of the rule may depend on interpretations not contained in the source material. The line between covered content and exempt material could prove as important as the headline obligation to label.
Campaigns outside Pennsylvania may still be affected
The reported law is not framed around where an advertisement is created. Instead, the account says disclosures may be required for ads expected to air in Pennsylvania even if they were made elsewhere. That can matter for campaigns designed for multiple markets. An advertiser that intends to reach Pennsylvania may have to account for the state’s rule even when its production, agency work or other distribution takes place outside the state.
For multi-state campaigns, that could encourage one version of an ad with a disclosure rather than separate versions for different jurisdictions. That is an implication of the reported geographic approach, not a requirement described in the supplied material. Whether businesses adopt a single standard, tailor campaigns by location, or avoid certain AI uses will depend on the law’s implementation and on their own assessment of the advertising involved.
The provision could also make the question of intended distribution important. The supplied account says the requirement can apply where an ad is expected to air in Pennsylvania, but it provides no further detail on how that expectation is established for advertisements that circulate through broad digital channels or across state lines. The material does not describe enforcement procedures, penalties, an effective date, or a state agency’s role.
Nor does the available reporting describe guidance for small businesses, advertising agencies, technology providers or platforms that may be involved in producing and distributing ads. A restaurant flyer and a national commercial may have very different production processes, yet the account presents both as potentially within the law’s scope. How the same statutory concepts operate across those settings is unresolved on the supplied record.
The policy goal is clearer than the operating rules
Pennsylvania’s action, as described, is a consumer-disclosure measure rather than a general prohibition on AI in advertising. It does not appear from the supplied material to ban the use of AI-generated images, voices, video or text. Its approach is to require a notice in specified circumstances, while allowing exclusions for material such as parody and plainly fictional work.
That structure puts transparency at the center of the policy. But transparency depends on a disclosure that is intelligible, noticeable and required in the cases lawmakers intended to reach. The account made available for this article does not reveal what form of label will satisfy the law, how an advertiser will document compliance, or how close cases will be resolved. Until those issues are clearer, statements about the law’s eventual impact should be cautious.
The report also does not provide the bill text, official signing materials, implementing guidance or an explanation from the office of State Senator Nick Pisciottano, who introduced the measure. The Verge said that office did not immediately respond to a request for comment. As a result, the account establishes a reported enactment and outlines the reported framework, but leaves important operational details unconfirmed.
This report has not been independently corroborated. It relies on a single secondary-source account and the limited claims supplied for publication. The core reported facts are that Shapiro signed the act, that certain AI-generated or significantly AI-modified ads will require same-format disclosures, and that the law includes specified exceptions; the exact legal boundaries and implementation remain uncertain.
For further context on this subject, see Anthropic says it will commit 5% of shares to a public-benefit fund.
Reporting notes
What is confirmed: The reported statute ties disclosure to AI content that creates a false consumer perception about a product or service.
Why this matters: The rule could affect ads distributed into Pennsylvania across image, video, audio and text formats, including campaigns made elsewhere.
What remains unclear: The available account does not detail effective dates, enforcement, label specifications or how key terms will be interpreted. This report is based on one source and has not been independently corroborated.
