Three jurisdictions covering roughly 2 billion people now require advertisers and platforms to disclose artificially generated content, and the last starts applying in nine days. New York's synthetic performer rules took effect on June 9, 2026, India's amended intermediary rules on February 20, 2026, and Article 50 of the European Union's AI Act becomes applicable on August 2, 2026.

The three regimes were drafted separately, define their subject matter differently, and attach penalties that differ by four orders of magnitude. What they share is a compliance calendar that has now closed. For an advertiser running a campaign with a computer-generated presenter across all three markets, the practical question is no longer whether disclosure is required but which of three incompatible definitions the creative must satisfy.

New York's law reached effect 180 days after signing

Governor Kathy Hochul announced on June 9, 2026 that New York's synthetic performer disclosure law had entered into force, exactly 180 days after it was signed as Chapter 617 of the Laws of 2025 on December 11, 2025.

"In New York, we are setting the rules of the road instead of letting AI run the show," Hochul said in the announcement. "Requiring simple, honest disclosure when an ad uses synthetic performers protects consumers, respects our creative workforce and keeps New York at the forefront of responsible innovation."

The measure amends section 396-b of New York's General Business Law, a provision originally added by chapter 1031 of the laws of 1965. The operative language sits in a new subdivision 3. Any person engaged in the business of dealing in property or a service who, for a commercial purpose, produces or creates an advertisement must conspicuously disclose that a synthetic performer appears in it, where that person has actual knowledge.

Two elements of that construction carry weight. The obligation attaches to the producer or creator, not to the medium that carries the advertisement. And the actual knowledge condition places the duty on the party positioned to know how the creative was assembled.

The penalty schedule is modest by regulatory standards: one thousand dollars for a first violation, five thousand for any subsequent one. The bill passed the New York Senate on June 13, 2025 by a floor vote of 59 in favour and none against, with four members excused.

The definition turns on non-recognisability

New York defines a synthetic performer as a digitally created asset created, reproduced, or modified by computer, using generative artificial intelligence or a software algorithm, that is intended to create the impression that the asset is engaging in an audiovisual or visual performance of a human performer who is not recognisable as any identifiable natural performer.

That closing clause matters. The statute targets fabricated humans who resemble nobody in particular. An AI-generated likeness of an identifiable person falls outside it, addressed instead by New York's civil rights law, which subdivision 5 expressly preserves.

State Senator Michael Gianaris, the bill's sponsor, framed the measure around labour protection. New York's performers will be better protected from their likenesses being deceptively replaced by artificial intelligence, Gianaris said, adding that the state can now hold accountable those who use the tools to profit at the expense of real talent.

Assemblymember Linda B. Rosenthal, who carried the companion measure in the Assembly, put the consumer case. "As companies rely on new marketing strategies to sell their products and services, consumers have a right to know if the person featured in those ads is real or fake and our first-in-the-nation law will ensure that happens," Rosenthal said.

SAG-AFTRA, which lobbied for the measure, characterised the outcome through its Chief Labor Policy Officer and New York Local Executive Director Rebecca Damon. "The law's enforceable protections mitigate performance replacement, prevent consumer deception, and affirm the continued value of human performance," Damon said.

Four carve-outs narrow the reach considerably

The New York statute contains exemptions that substantially limit which campaigns fall inside it.

Audio advertisements are excluded outright under subdivision 7(a). A synthetic voice reading a radio spot triggers no disclosure obligation in New York, a gap the European and Indian regimes both close. Language translation is excluded under subdivision 7(b), where the use of artificial intelligence solely involves translating a human performer's speech.

Expressive works receive a broad exemption under subdivision 4. Advertisements and promotional materials for motion pictures, television programs, streaming content, documentaries, video games, and similar audiovisual works fall outside the requirement, provided the use of a synthetic performer is consistent with its use in the underlying work. A trailer for a film featuring a synthetic character needs no separate disclosure.

Media carriers are shielded under subdivision 8. Newspapers, magazines, television networks and stations, streaming services, cable television systems, billboards, and transit advertisements bear no liability for publishing or disseminating a non-compliant advertisement. Subdivision 6 separately preserves the Section 230 position for interactive computer services.

The net effect places the entire compliance burden on the advertiser and the production company, with no residual duty on the platform or publisher. That allocation matches the structure Google adopted in July when it shifted AI ad labeling liability entirely to advertisers.

India's rules have been running since February

India's framework arrived first and reaches furthest into platform architecture. The Ministry of Electronics and Information Technology published the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 in the Gazette of India on February 10, 2026 under notification G.S.R. 120(E). They took effect on February 20, 2026, and compressed the content takedown window from thirty-six hours to three.

India defines synthetically generated information as audio, visual or audio-visual information which is artificially or algorithmically created, generated, modified or altered using a computer resource, in a manner that such information appears to be real, authentic or true and depicts or portrays any individual or event in a manner that is, or is likely to be perceived as indistinguishable from a natural person or real-world event.

The definition is broader than New York's in one respect and narrower in another. It captures audio, which New York exempts, and does not require that the depicted person be unrecognisable, so deepfakes of identifiable individuals fall inside it. But it excludes pure text, meaning a chatbot-generated article containing false claims is not synthetically generated information, though it may still be unlawful under other provisions, according to the ministry's FAQ dated February 10, 2026.

Three categories of exclusion were written into the definition itself. Routine or good-faith editing, formatting, enhancement, technical correction, colour adjustment, noise reduction, transcription, or compression falls outside it where it does not materially alter, distort, or misrepresent the substance, context, or meaning of the underlying content. Routine good-faith creation of documents, presentations, educational materials and research outputs is excluded, provided no false document or false electronic record results. And the use of computer resources solely for improving accessibility, clarity, translation, description or searchability is excluded where no material part of the underlying content is manipulated.

The ministry's FAQ works through the boundary with examples. Increasing brightness on a photograph, removing background noise, adding subtitles, and blurring number plates for privacy all fall outside. Using AI tools to generate fake certificates, forged identity documents or fabricated electronic records does not, and may be treated as unlawful synthetic content.

Labelling obligations attach to the tool, then to the platform

New rule 3(3) imposes a two-part duty on any intermediary offering a computer resource capable of generating synthetic content. First, the intermediary must deploy reasonable and appropriate technical measures, including automated tools, to prevent users from generating synthetic content that violates Indian law, a list naming the Bharatiya Nyaya Sanhita, 2023, the Protection of Children from Sexual Offences Act, 2012 and the Explosive Substances Act, 1908, and covering non-consensual intimate imagery, false documents, explosives and arms content, and false depictions of natural persons or real-world events.

Second, all other synthetic content must be labelled in a manner ensuring prominent visibility in the visual display, easily noticeable and adequately perceivable, or for audio content, through a prominently prefixed audio disclosure. The ministry's FAQ gives the example of a synthetic voice narration carrying a spoken disclosure such as "This audio is synthetically generated" at the beginning.

The rule also requires embedded permanent metadata or another provenance mechanism, including a unique identifier, to the extent technically feasible, identifying the computer resource used to generate the content. Rule 3(3)(b) forbids the intermediary from enabling the modification, suppression or removal of that label, metadata or identifier. The FAQ states the consequence plainly: an intermediary should not offer "remove watermark", "export without metadata", or similar functionality.

An earlier draft would have required visible labels covering ten percent of a display's surface area or duration. The final text dropped that specification in favour of the prominence standard, giving platforms greater implementation flexibility.

New rule 4(1A) adds an ex-ante layer for significant social media intermediaries. Before displaying, uploading or publishing any information, such a platform must require users to declare whether the content is synthetically generated, deploy technical measures to verify that declaration, and apply a clear and prominent label where synthetic origin is confirmed. A platform that knowingly permits, promotes or fails to act on non-compliant synthetic content is deemed to have failed its due diligence obligation, which under section 79 of the IT Act is the condition on which safe harbour rests.

Article 50 applies in nine days

The European regime becomes applicable on August 2, 2026. The European Commission published its implementation guidelines as Communication C(2026) 5054 final on July 20, 2026, alongside a finalised Code of Practice on Transparency of AI-Generated Content, fixing the technical detail before the obligations bind.

Article 50 of Regulation (EU) 2024/1689 splits its duties between providers and deployers. Providers of systems intended to interact directly with natural persons must design them so that people are informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant and circumspect person. Providers of systems generating synthetic audio, image, video or text content, including general-purpose systems, must ensure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated.

The deployer duty is the one that reaches advertisers. Under Article 50(4), deployers of an AI system that generates or manipulates image, audio or video content constituting a deep fake must disclose that the content has been artificially generated or manipulated. Where the content forms part of an evidently artistic, creative, satirical, fictional or analogous work, the obligation narrows to disclosure in a manner that does not hamper the display or enjoyment of the work.

That narrowing does not extend to advertising. Commission guidance has treated persuasive commercial content as outside the lighter artistic regime, with an AI-manipulated video using synthetic humans to sell a product listed as content that does not constitute an artistic work.

A second deployer duty covers AI-generated text published to inform the public on matters of public interest, unless the content has undergone human review or editorial control and a natural or legal person holds editorial responsibility for it.

Article 50(2) contains an exemption with no analogue in New York: the marking obligation does not apply where AI systems perform an assistive function for standard editing or do not substantially alter the input data or its semantics. That covers roughly the same territory as India's good-faith editing exclusion, reached by a different drafting route.

Enforcement runs through national market surveillance authorities, the AI Office, and the European Data Protection Supervisor where EU institutions are the provider or deployer. Under Article 99(4)(g), non-compliance with Article 50 attracts administrative fines of up to 15 million euros or, for an undertaking, up to 3 percent of total worldwide annual turnover for the preceding financial year, whichever is higher. For SMEs including start-ups, the applicable cap is whichever is lower.

The August 2 date has survived one attempt at revision. Digital Omnibus negotiations moved compliance dates for high-risk systems without touching Article 50. Staged deadlines follow: systems already on the market have until 2 December 2026 to bring machine-readable marking into conformity, and a watermark-detection interoperability requirement lands on 2 February 2027.

Three definitions, one creative asset

The compliance problem is definitional rather than procedural. An advertisement featuring a computer-generated presenter who resembles no real person, running in all three markets, must satisfy three tests written to different specifications. New York asks whether the asset creates the impression of a performance by a human performer not recognisable as any identifiable natural performer, and exempts audio entirely. India asks whether the content appears real, authentic or true and is likely to be perceived as indistinguishable from a natural person or real-world event, and covers audio. The European Union asks whether the content constitutes a deep fake.

The penalty ranges diverge as sharply. A repeat violation in New York costs $5,000. A comparable failure in Europe reaches 15 million euros or 3 percent of worldwide turnover. India attaches no direct fine, using instead the loss of intermediary safe harbour, a sanction with no fixed ceiling.

Enforcement targets differ too. New York names the producer or creator. India names the intermediary offering the generation tool and the platform publishing the output. The European Union splits the duty between provider and deployer.

Google's response has been to build a single control and disclaim its legal sufficiency. The company introduced an AI label setting across Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center and Ads Editor in a changelog dated July 9, 2026, with a visible on-ad overlay restricted to campaigns targeting the European Union, India and New York. That the same three jurisdictions appear in the company's own documentation confirms the convergence, as PPC Land reported when the five-product rollout was announced. The disclaimer attached to that setting is unambiguous: its use does not guarantee compliance with specific regulations.

A conflict has already surfaced inside Google's own policies. The labeling regime requires visible text on creative, while image asset requirements restrict text overlays, a tension documented when the platform banned blurry image assets and cut eligibility to 60-day accounts. Automatic image-enhancement features that crop assets can also remove a custom label from frame.

The labels arrive into an audience already sceptical. IAB research surveying 505 US Gen Z and Millennial consumers alongside 104 industry executives between October 2025 and January 2026 found 39 percent of Gen Z respondents reporting negative sentiment toward AI ads, against 20 percent of Millennials, with the perception gap between consumers and executives widening from 32 points in 2024 to 37 points in 2026. Adoption is no longer marginal either: XR found 26 percent of respondents already using AI digital replicas and 22 percent using AI synthetic talent, with full-service agencies leading at 31 percent.

New York's Office of Digital Innovation, Governance, Integrity, and Trust, announced alongside the June 9 statement, will initially focus on large frontier AI developers rather than advertisers. Enforcement of section 396-b subdivision 3 remains untested, with no public action recorded since June 9.

Timeline

Summary

Who: New York State, under Governor Kathy Hochul with sponsors Senator Michael Gianaris and Assemblymember Linda B. Rosenthal; India's Ministry of Electronics and Information Technology; and the European Commission, alongside advertisers, agencies, production companies and platforms operating in those markets.

What: Three separate AI disclosure regimes now govern advertising and platform content. New York requires conspicuous disclosure of synthetic performers in advertisements, carrying a $1,000 civil penalty for a first violation and $5,000 for subsequent violations. India requires intermediaries to prevent unlawful synthetic content, label permissible synthetic content prominently, embed tamper-resistant provenance metadata, and, for significant social media intermediaries, collect and verify user declarations before publication. The European Union requires providers to mark synthetic outputs machine-readably and deployers to disclose deep fakes, with fines up to 15 million euros or 3 percent of worldwide turnover.

When: India's rules were published on February 10, 2026 and took effect on February 20, 2026. New York's law was signed on December 11, 2025 and took effect on June 9, 2026. Article 50 of the EU AI Act becomes applicable on August 2, 2026, with guidelines published on July 20, 2026 and further deadlines on December 2, 2026 and February 2, 2027.

Where: New York State; India, covering intermediaries serving its internet users regardless of incorporation; and the European Union, applying to providers and deployers placing AI systems or content on the European market.

Why: All three regimes cite consumer deception as the driving concern, with New York adding performer displacement and India adding impersonation, electoral manipulation and non-consensual imagery. The definitional divergence between the three creates a practical difficulty for cross-border campaigns, since a single creative asset can fall inside one regime, outside another, and into a different enforcement structure in the third.