Pennsylvania Sends an AI Ad Disclosure Bill to the Governor

Pennsylvania’s House voted 163-40 on October 7 to pass a bill requiring disclosures for certain deceptive uses of AI in advertising. The legislative record for Senate Bill 806 shows final passage in both chambers. As of the October 8 record checked for this article, a governor’s signature hadn’t been recorded. Signing would start a 60-day wait before the act takes effect.

For a creative team, the immediate job is to identify which generated assets make a claim about what the buyer will receive. Then decide how that claim and any required label travel together. A label approved in a design file won’t help if the exported ad crops it out.

State Senator Nick Pisciottano, the bill’s prime sponsor, said in the October 7 announcement:

“Consumers deserve to know if AI is being used to create a false impression about a product or service.”

The proposal puts the consumer’s impression of the finished advertisement at the center of review. That gives the person approving creative a more useful starting point than a count of how many production steps involved AI.

What would trigger disclosure

The bill’s current text covers knowing, reckless, or willful creation and dissemination of certain ads using AI-generated or significantly AI-modified content. The trigger is a false impression about a product or service’s nature, origin, or authenticity. Covered ads would need a clear disclosure that AI generated or modified the content. Text-only content, evident fiction, parody, satire, commentary, and criticism are exceptions.

A disclosure would have to appear when consumers first encounter the content, use the same medium, and remain attached or be exceptionally difficult to remove where technically feasible. Visibility and audibility matter: size, contrast, position, duration, and language must make the notice understandable. The text assigns enforcement to the attorney general or an appropriate district attorney.

Those provisions suggest a practical review of the delivered asset. Take a hypothetical product image showing a feature the actual product lacks. The reviewer needs to examine the impression created by that image, with counsel deciding whether the legal test applies. An AI-use label shouldn’t be treated as approval to publish a misleading claim: the bill explicitly preserves other legal recourse.

For software marketers, the equivalent review concerns depicted screens and capabilities. An illustration of a proposed interface needs a different brief from a screenshot presented as the working product. These are editorial examples of the approval questions, not findings about specific campaigns or predictions about enforcement.

The lawyer’s implementation question

Libby O’Neill is deputy chair of Advertising, Marketing & Promotions at Loeb & Loeb LLP. Their account published July 17 described the firm’s July 14 advertising roundtable. The discussion included Pennsylvania’s pending measures, brand counsel, and agency lawyers. It wasn’t a response to this week’s passage.

Writing about the group’s training discussion, O’Neill said:

“short, scenario-based materials outperform lengthy policy memos”

That observation is practitioner experience, not a measured training experiment. It points to a workable format for the creative desk: a few approved and rejected examples, with the reason for each decision beside the asset. Loeb advises clients on AI governance and disclosure, so the firm has a commercial interest in the issue.

O’Neill also described a discussion about uniform labeling across platforms versus notices tailored to individual channels. Participants leaned toward more disclosure while expressing concern about consumer skepticism. The unresolved choice belongs in a campaign brief: who decides the label, and who checks whether it remains legible in each placement?

The House committee’s October 5 announcement framed the measure around consumers being able to judge what they’re seeing when buying goods. That concern reaches the final export, where the audience encounters the product claim without seeing the team’s prompts or production notes.

Our reporting on trust in AI product imagery covers whether a picture accurately represents the product. Generated artist signatures raise a related question: whether a mark implies a person made or endorsed it. This bill adds a specific disclosure proposal to the broader AI content and brand approval work.

Three changes to the creative handoff

1. Attach the product evidence. Put the final image beside a product photo, working screenshot, or approved specification. Record what the generated asset depicts, and flag departures for the creative approver and counsel. The production tool’s name alone won’t explain the claim.
2. Review the placement export. Test the label in the actual crop, playback length, or audio treatment. Assign someone to check the version the customer sees. Keep that export with the approval record so a later resize doesn’t silently substitute a different asset.
3. Keep the legal status dated. SB 806 has passed the legislature; the signature and calendar effective date remain unverified in the record checked here. Recheck enactment before treating the 60-day window as an active deadline, and use campaign-specific legal review to determine coverage.