Oct 25 edition/Reporting & analysis
PolicyBusinessMultimodal

PolicyLaw, regulation & governance

Pennsylvania enacts law requiring labels on misleading AI-generated ads

Pennsylvania Gov. Josh Shapiro signed SB 806, requiring labels when AI-generated or significantly modified content in an ad creates a false impression for consumers. The rule covers ads shown in the state wherever they were made, with fines up to $3,000 per violation.

From the The Verge · AI article of October 24, 2026 and cited sources: Shapiro signed the Artificial Intelligence Deceptive Advertising Act on October 22, 2026, after a 48–2 Senate vote and a 163–40 House vote. [1] [3] [6] [8]

SOURCES[1] Pennsylvania will require labels on AI-generated ads theverge.com[3] River Reporter: NY, PA regulate AI in advertising riverreporter.com[6] Sen. Pisciottano press release pasenatornick.com

AI-assisted, not human-edited. Methodology Corrections Who owns this site

KEY FACTS
What happened
Shapiro signed the Artificial Intelligence Deceptive Advertising Act on October 22, 2026, after a 48–2 Senate vote and a 163–40 House vote. [1] [3] [6] [8]
Original source
The Verge · AI article, independent perspective
Brief by
AT Pulse research desk AI-assisted; publisher and contact on the desk page
When
October 24, 2026 article published
Evidence
10 sources across 10 sites
48/100RisingHow it’s measured

YouTube video · 205.9K views · 4.9K likes

Reach blends public engagement (views, likes, comments, shares, stars) from the platforms where the story or its original source lives: YouTube, Hacker News, GitHub, Substack and similar. A video and the article written from its transcript share one score, taken from the video. Platforms with no data are left out rather than counted as zero, and the age of a story never lowers its score.

Illustration of ad broken by a gavel
Image: The Verge — Original article ↗
THE CORE IDEAS4 TAKEAWAYS
01

Shapiro signed the Artificial Intelligence Deceptive Advertising Act on October 22, 2026, after a 48–2 Senate vote and a 163–40 House vote. It takes effect 60 days after signing. A label is needed only when synthetic content misleads consumers about a product's nature, origin or authenticity. Fines reach $3,000 per violation, and the law applies to any covered ad shown in Pennsylvania, including ads made out of state. [1] [3] [6] [8]

02

The law sets rules for how labels look and sound, not for how AI content is detected. A label must appear when the synthetic content is first shown and must use the ad's own medium: spoken in audio, on screen in video. It must be audible or readable for long enough, and where technically feasible it should be very hard to remove. The reviewed sources name no watermarking or provenance standard and give no technical definition of 'significantly modified.' [5] [1]

03

Accounts of the law disagree on basic points. In the linked video, the sponsoring senator's segment says only AI companies would be penalized. A law-firm analysis of the bill text says advertisers carry primary liability and the penalties are civil. Sources also split on whether text-only ads are covered or exempt. [2] [5] [3]

04

Only the state Attorney General and county district attorneys can enforce the law; consumers cannot sue under it. NetChoice, a tech trade group, argued that existing laws already ban deceptive ads and that the label rules could overwhelm short videos, small display ads and audio spots. [5] [7]

WHY IT MATTERS

The law's text reaches national campaigns that air in Pennsylvania and sets specific label formats. As a result, advertisers and creative-tool vendors will likely need labels built for each medium that are hard to remove.

Read the full assessment

Those labels may still not satisfy New York's or California's different rules.

Research is finished and I'm writing up the dossier now.

Executive brief

Pennsylvania's new law on AI in ads reaches beyond the state's borders. Any covered ad shown in Pennsylvania must carry an AI label, wherever it was made, so national campaigns that air there have to comply (The Verge). Josh Shapiro signed SB 806, the Artificial Intelligence Deceptive Advertising Act, on October 22, 2026. A label is required only when synthetic content gives consumers a false impression of a product. Fines go up to $3,000 per violation (River Reporter). Coverage disagrees on two basic points: whether AI companies or advertisers are liable, and whether text is covered.

Read the full section

Pennsylvania's new law on AI in ads reaches beyond the state's borders. Any covered ad shown in Pennsylvania must carry an AI label, wherever it was made, so national campaigns that air there have to comply (The Verge). Gov. Josh Shapiro signed SB 806, the Artificial Intelligence Deceptive Advertising Act, on October 22, 2026. A label is required only when synthetic content gives consumers a false impression of a product. Fines go up to $3,000 per violation (River Reporter). Coverage disagrees on two basic points: whether AI companies or advertisers are liable, and whether text is covered.

What changed and event timeline

  1. Committee advances SB 806

    The Senate Communications and Technology Committee approved the bill unanimously. A similar House bill, HB 95, cleared its committee on May 5 ().

  2. Senate passes the bill 48–2

    The Senate vote sent the bill to the House (;).

  3. NetChoice objects

    The tech trade group wrote to lawmakers that existing laws already ban deceptive ads. It also said the formatting rules could overwhelm short videos, small display ads and audio spots ().

  4. House gives final approval, 163–40

    The bill went to Gov. Shapiro ().

  5. Shapiro signs

    The law takes effect after 60 days, around December 21, 2026 (;).

Capabilities and access

This is a law, not an AI model, so there is no model or version involved.

Read the full section

This is a law, not an AI model, so there is no model or version involved.

  • What it covers: The law defines "synthetic content" as images, video, audio or text "significantly modified or generated" by AI (Ginsburg Law Group).
  • When a label is required: An advertiser who knowingly, recklessly or willfully runs an ad whose synthetic content creates a false impression about "nature, origin or authenticity" must disclose it (River Reporter).
  • Exemptions:
  • content that is evidently fictional
  • parody, satire, commentary or criticism
  • text-only content (River Reporter)
  • political ads, according to CBS (CBS Pittsburgh)

Technical analysis for researchers and developers

The law sets rules for how a label must appear, but not for how AI content is detected (Ginsburg Law Group): The reviewed sources do not name a watermarking or provenance standard. They also do not define "significantly modified" or "false consumer perception" in technical terms (The Verge). Builders of generation pipelines will need their own internal thresholds until enforcement shows how these terms are read.

Read the full section

The law sets rules for how a label must appear, but not for how AI content is detected (Ginsburg Law Group):

  • Timing: The label must appear when the synthetic content is first shown.
  • Medium: It must use the ad's own medium: spoken for audio, on screen for video.
  • Readability: Audio labels need enough volume and a slow enough pace to be heard. On-screen labels must stay up long enough to read.
  • Durability: Where technically feasible, the label should be "permanent or extraordinarily difficult to remove" from the content.

The reviewed sources do not name a watermarking or provenance standard. They also do not define "significantly modified" or "false consumer perception" in technical terms (The Verge). Builders of generation pipelines will need their own internal thresholds until enforcement shows how these terms are read.

Claims and evidence

  • Out-of-state ads are covered. The Verge reports this (The Verge), and Sen. Pisciottano's segment says the same (video 01:19).
  • Who is liable is disputed. AI platforms are not directly targeted (Ginsburg Law Group).
  • Whether text is covered is disputed. The Verge and the senator say text ads are covered (video 01:07).
Read the full section
  • Out-of-state ads are covered. The Verge reports this (The Verge), and Sen. Pisciottano's segment says the same (video 01:19).
  • Who is liable is disputed. The KDKA segment says only AI companies, not users, face fines or even jail time (video 01:46). A law-firm summary of the bill text says advertisers carry primary liability and the penalties are civil. AI platforms are not directly targeted (Ginsburg Law Group).
  • Whether text is covered is disputed. The Verge and the senator say text ads are covered (video 01:07). Other summaries say text-only ads are exempt (River Reporter).
  • Scams are the stated motive. The senator says groups representing older consumers reported people being scammed by fake AI ads (video 01:34). This is his account; the reviewed sources give no data on it.

Context and prior work

  • New York: A law effective December 11, 2025 requires disclosure of AI deepfakes that imitate real people in ads.
  • Newsom recently signed a rule requiring disclosure of "synthetic" digital performers in ads (MediaPost).
  • Scope difference: Pennsylvania's test covers any deceptive synthetic content, not only imitations of real people.
Read the full section
  • New York: A law effective December 11, 2025 requires disclosure of AI deepfakes that imitate real people in ads. It exempts animated characters and audio-only ads. Fines are $1,000 for a first offense and $5,000 after that (River Reporter).
  • California: Gov. Newsom recently signed a rule requiring disclosure of "synthetic" digital performers in ads (MediaPost).
  • Scope difference: Pennsylvania's test covers any deceptive synthetic content, not only imitations of real people.
  • Industry views: The PA Chamber supported the bill with amendments in May (PA Chamber memo). NetChoice opposed it.

Limitations, safety and contested findings

  • Vague terms: "Significantly modified," "evidently fictional" and "false" are subjective, and the law does not say who decides (The Verge).
  • Narrow trigger: A label is needed only if the content misleads.
  • Limited enforcement: There is no private right of action. Only the Attorney General and county district attorneys can bring cases (Ginsburg Law Group).
Read the full section
  • Vague terms: "Significantly modified," "evidently fictional" and "false" are subjective, and the law does not say who decides (The Verge).
  • Narrow trigger: A label is needed only if the content misleads. AI ads that do not mislead need no label, which differs from how the senator described the law on air (video 00:59).
  • Limited enforcement: There is no private right of action. Only the Attorney General and county district attorneys can bring cases (Ginsburg Law Group).
  • Labels don't protect against other claims: Disclosing AI use does not shield an advertiser from other consumer-protection laws (Ginsburg Law Group).
  • Burden on small formats: NetChoice argues the label rules could overwhelm short videos, small display ads and audio spots (MediaPost).

Business and practitioner implications

  • Compliance date: National advertisers and agencies need labeling workflows for anything airing in Pennsylvania by about December 21, 2026.
  • Liability: Until the AI-company-versus-advertiser question is settled, brands should assume they carry the risk, which follows the bill-text reading.
  • Creative tools: Vendors may get requests for built-in labels that are hard to remove and fit each medium.
Read the full section
  • Compliance date: National advertisers and agencies need labeling workflows for anything airing in Pennsylvania by about December 21, 2026.
  • Liability: Until the AI-company-versus-advertiser question is settled, brands should assume they carry the risk, which follows the bill-text reading.
  • Creative tools: Vendors may get requests for built-in labels that are hard to remove and fit each medium.
  • Multi-state planning: The New York, California and Pennsylvania rules differ in what they cover and what they exempt, so one label policy may not satisfy all three.
FOLLOW THE EVIDENCE

The source trail.

Sources (10)
A LITTLE LESS NOISE. A LOT MORE CONTEXT.

Stay curious.
Follow the evidence.

Independent perspectives, the original sources, and room for the questions that don't have easy answers.

How we build the brief
Connect with us

Find us where you already read.