States Are Quietly Writing an AI Truth-in-Advertising Code While Congress Stalls
On October 7, Pennsylvania's House of Representatives passed SB 806 by a 163-40 margin, sending to Governor Josh Shapiro a bill that bans advertisers from using AI to falsely shape how consumers perceive a product's or service's nature, origin, or authenticity unless they provide a "clear and conspicuous disclosure." The bill cleared the Senate in June 48-2. Those are not the vote tallies Washington produces on anything touching artificial intelligence this year.
That's the point. While Congress's AI accountability bills pile up without floor votes, a narrower category of state legislation — require disclosure when AI shapes an advertisement, don't touch model training or liability — is actually becoming law, and doing so with bipartisan margins that broader AI safety bills never get.
California already moved twice
Pennsylvania is catching up to California, which signed two AI-advertising disclosure statutes in the same September window it signed its broader AI employment and auditing package. SB 1050 (Chapter 246, Statutes of 2026, signed September 16) requires disclosure whenever an ad prominently features a "synthetic performer" — a generative-AI figure, voice, or likeness simulating a human who isn't a real, identifiable person — with carve-outs for movies, TV, streaming, games, and documentaries. AB 2025 (Chapter 572, signed September 27) requires real estate and rental listings to disclose when photos have been digitally altered, name the tool used, and make the original, unaltered image available on request.
New York got there first: its own synthetic-performer disclosure law (A 8887/S 8420) took effect in June. Utah, Maryland (SB 827), and New Jersey (A 5607) are now working similar territory from the other direction — restricting AI chatbots from using logged conversations for targeted advertising — while Kentucky (HB 641) and Louisiana (SB 5) have floated bans on mental-health chatbots advertising products to the people using them.
A truth-in-advertising extension is a much easier sell than a model-liability bill: it doesn't ask legislators to referee what AI companies can build, only what advertisers can show people.
Why this lane clears votes others can't
The pattern across all five bills is the same: disclosure, not restriction. None of them limit what AI tools can generate; they require a label when a human might otherwise be deceived about what's real. That framing lets lawmakers plug AI into existing false-advertising and consumer-protection statutes — Pennsylvania's bill amends unfair-competition law directly — rather than writing new AI-specific liability regimes from scratch. It's also why the vote counts look nothing like the 50-50 fights over frontier-model safety bills: disclosure mandates poll well with both parties' consumer-protection instincts, and ad-industry lobbying has focused on carve-outs (expressive works, political speech) rather than killing the bills outright.
For decision-makers, the near-term exposure sits with anyone running AI-generated or AI-altered creative at scale: retail media networks, ad agencies and the platforms serving their inventory (Meta, YouTube, TikTok), proptech brokerages (Zillow, Redfin, Compass) using AI-staged listing photos, and any brand using AI dubbing or synthetic spokespeople in paid media. California's two laws, signed without urgency clauses, take effect under the state's standard January 1, 2027 cycle — which means holiday-season Q4 2026 campaigns are the last ones built under the old rules before compliance reviews become mandatory in the country's largest ad market. Pennsylvania's SB 806 will set the next data point: if Shapiro signs it, that's four states (PA, CA, NY, and whichever of UT/MD/NJ lands first) with live AI-ad-disclosure law within a single year, which is the kind of critical mass that tends to pull a de facto national compliance standard behind it even without a federal law.