Disclosure rules

AI disclosure laws: do you have to disclose AI-generated content?

No single rule requires labeling all AI content. Instead, a set of EU, state and federal laws each reach chatbots, synthetic media, ads or elections.

SD Frivolous Editorial · Published Nov 12, 2026 · 12 sources

Short answer

Sometimes, depending on what you made and where it reaches people. As of November 9, 2026, no US law requires a label on all AI-generated content. Disclosure is required for chatbots that could pass as human (EU and several states), for generative AI image, video and audio in the EU and California, for synthetic performers in New York ads, and for AI political ads in about 30 states.

Editorial illustration for AI disclosure laws

At a glance
General US labeling law	None as of November 9, 2026; duties come from specific state laws and the FTC Act
EU rule	AI Act Article 50, applies from August 2, 2026; marking grace to December 2, 2026 for systems already on the market
California provenance rule	AI Transparency Act (SB 942, as amended), provider duties operative August 2, 2026
Chatbot disclosure laws	California (BOT Act, SB 243, AB 1609 from 2027), New York, Utah, Maine
Advertising	New York synthetic performer law, GBL § 396-b, in effect since June 9, 2026
Elections	30 states with election deepfake laws as of May 2026 (Public Citizen)
Top EU fine	Up to EUR 15 million or 3% of worldwide turnover for Article 50 breaches

Do you have to disclose AI-generated content?

AI disclosure means telling the people who see or interact with something that AI produced it or is running it. The law does not treat that as one duty. It attaches disclosure to four situations: a machine talking to a person, synthetic media that could be mistaken for real, advertising, and elections. Outside those situations, using AI to draft a report, an email or a product description generally triggers no statutory label in the United States.

The answer therefore turns on three questions. First, what is the output: a live chatbot, an image or video, text, or an ad? Second, who is the audience: EU residents, Californians, New Yorkers, voters? Third, what is your role: the company that built the model, the business deploying it, or an advertiser? The European Union regulates by role, putting marking duties on providers and labeling duties on deployers under Article 50 of the AI Act, Regulation (EU) 2024/1689. US states regulate by use case, one statute at a time.

Federal law fills the gaps through deception. The FTC Act bars unfair or deceptive practices, and the FTC's rule on consumer reviews, which took effect October 21, 2024, bans fake reviews and testimonials, including AI-generated ones (Alston & Bird). There is no federal statute requiring a label on AI content as such.

AI disclosure requirements by law

The table lists the main disclosure rules in force or enacted as of November 9, 2026, with the source for each row. It covers the US and the EU; South Korea's labeling duty is noted further down.

Law	What must be disclosed	Who must disclose	Applies from	Key authority
EU AI Act, Article 50	That a user is talking to AI; machine-readable marking of synthetic audio, image, video and text; labels on deepfakes and AI text on public-interest matters	Providers (chatbots, marking) and deployers (labels)	Aug. 2, 2026; marking grace to Dec. 2, 2026 for systems already on the market	EUR-Lex; Gibson Dunn
California AI Transparency Act (SB 942, AB 853, SB 1000)	Latent provenance data in AI image, video and audio; free verification tool	Generative AI providers; platforms from 2027	Aug. 2, 2026	Morgan Lewis
California BOT Act, Bus. & Prof. Code § 17941	That a bot is a bot, when used to mislead people to drive a sale or influence a vote	Anyone using a bot online with Californians	Enacted 2018	§ 17941 text
California SB 243	That a companion chatbot is AI where a reasonable person could be misled; suitability warning for minors	Companion chatbot operators	Jan. 1, 2026	Gunderson Dettmer
California AB 1609	That a customer service chatbot is automated; path to a human agent	Businesses with over $500 million revenue	Jan. 1, 2027	Manatt
New York AI companion law, GBL Article 47	That the user is not talking to a human, at the start and at least every three hours	AI companion operators	Nov. 5, 2025	FPF; Morrison Foerster
Utah Artificial Intelligence Policy Act	Use of generative AI, on a clear and unambiguous request; up front in high-risk regulated-occupation interactions	Businesses using generative AI with consumers	May 1, 2024; amended May 7, 2025	Davis Polk
Maine LD 1727	That a chatbot is AI, where it may mislead a consumer into thinking it is human	Businesses using chatbots in commerce	2025	Maine House
New York synthetic performer law, GBL § 396-b	That an ad contains a synthetic performer	Advertisers with actual knowledge	June 9, 2026	Cooley
State election deepfake laws	Disclaimers on AI-altered political content, in most of these states	Varies: campaigns, committees or anyone distributing	Varies by state	Public Citizen tracker
FTC Act § 5 and the Consumer Reviews Rule	No AI label duty; bans deception, including AI-generated fake reviews	All businesses	Rule effective Oct. 21, 2024	Alston & Bird

EU AI Act Article 50: what must be disclosed in Europe

Article 50 of Regulation (EU) 2024/1689 is the closest thing to a general AI disclosure law. It sets four duties. Providers must design systems that interact directly with people so that users are told they are dealing with AI, unless that is obvious. Providers of systems that generate synthetic audio, image, video or text must mark the output in a machine-readable format that is detectable as artificially generated, as far as technically feasible. Deployers of emotion recognition or biometric categorisation systems must inform the people exposed. Deployers who publish deepfakes, or AI-generated text to inform the public on matters of public interest, must disclose that the content is artificial.

Article 50 applies from August 2, 2026. The 2026 Digital Omnibus left that date largely in place, adding only a grace period to December 2, 2026 for generative systems placed on the market before August 2, 2026 to meet the marking duty (Gibson Dunn). Breaches can draw fines of up to EUR 15 million or 3% of worldwide annual turnover under Article 99(4). The Act reaches providers placing systems on the EU market wherever they are established.

The text-labeling duty is narrower than often assumed. It applies to deployers publishing AI text to inform the public on matters of public interest, and the Article provides an exception where the text has undergone human review or editorial control and a person holds editorial responsibility. Creative, satirical and fictional deepfakes still need disclosure, but in a way that does not spoil the work. Our page on the EU AI Act content rules covers Article 50 in depth, and the EU AI Act overview covers the rest of the regulation.

California SB 942: provenance data instead of visible labels

California's AI Transparency Act, built on SB 942 (2024) and amended by AB 853 (2025) and SB 1000 (2026), is a provenance law rather than a labeling law. Since August 2, 2026, providers of generative AI systems that create image, video or audio must embed a latent disclosure in the file and offer a free tool that checks for it (Morgan Lewis). SB 1000, effective September 30, 2026, removed the one million user threshold and dropped the duty to offer a visible manifest label (Latham & Watkins).

Text is outside the act. Large online platforms and model hosting sites take on display and hosting duties on January 1, 2027, and capture device makers on January 1, 2028. Violations carry $5,000 per violation, each day a separate violation, enforced by the Attorney General and local prosecutors; there is no private right of action. The full breakdown is on our California AI Transparency Act page.

Do chatbots have to say they are AI? State chatbot disclosure laws

In several states, yes. The rules differ on when the notice is triggered.

California

California's BOT Act makes it unlawful to use a bot to communicate with a person in California online with intent to mislead them about its artificial identity, to drive a sale or influence a vote; a clear and conspicuous disclosure that it is a bot avoids liability (Bus. & Prof. Code § 17941). SB 243, in force since January 1, 2026, requires companion chatbot operators to give a clear notice that the chatbot is AI where a reasonable person could be misled, and to warn that companion chatbots may not be suitable for some minors (Gunderson Dettmer). From January 1, 2027, AB 1609 bars businesses with over $500 million in annual revenue from presenting customer service chatbots as human and requires a simple way to reach a person (Manatt).

New York

New York's AI companion law, General Business Law Article 47, took effect November 5, 2025. It requires notice at the start of an interaction and at least every three hours during an ongoing conversation, plus a protocol for suicidal ideation (Future of Privacy Forum). Only the Attorney General enforces it, with civil penalties of up to $15,000 per day (Morrison Foerster).

Utah and Maine

Utah's Artificial Intelligence Policy Act, in force since May 1, 2024, was narrowed in 2025: businesses must disclose generative AI use when a consumer makes a clear and unambiguous request, and regulated occupations must disclose up front only in high-risk interactions involving sensitive data and personalized advice. The Division of Consumer Protection can impose fines of $2,500 per violation (Davis Polk). Maine's LD 1727, signed in June 2025, bars using an AI chatbot in a way that may mislead consumers into believing they are dealing with a person unless they are clearly told, and makes a violation an unfair trade practice enforced by the Attorney General (Maine House Democrats).

Election and advertising disclosure

Political ads are where US disclosure law is densest. Maryland became the 30th state with election deepfake protections in May 2026, according to Public Citizen, which tracks the laws. Most of these laws require a disclaimer on materially deceptive AI-altered depictions of candidates within a window before an election rather than banning them; details, windows and penalties differ by state, so the tracker is the place to check a specific state.

Courts have limited how far these laws can go. In Kohls v. Bonta (E.D. Cal.), a federal judge in August 2025 struck down California's AB 2839, which targeted deceptive election content, finding it discriminated based on content and viewpoint (Bloomberg Law). California's separate disclosure rule for political committees, AB 2355, requires a disclaimer on qualified political ads with AI-generated or substantially altered content (Washington University School of Law).

Commercial advertising has one dedicated statute. New York's synthetic performer law, General Business Law § 396-b, has applied since June 9, 2026. Anyone who produces an ad and has actual knowledge that it includes a synthetic performer, a digitally created asset meant to look like a human performer who is not an identifiable person, must disclose it conspicuously. Penalties are $1,000 for a first violation and $5,000 for each later one, and ads for films, shows and games are exempt where the use matches the work (Cooley). Elsewhere, an undisclosed AI ad is judged under ordinary false advertising and FTC deception law.

What is AI disclosure in practice?

Compliance usually takes one of two forms: a technical label attached to the content, or a statement made to the people affected.

AI transparency labels

Technical labels are machine-readable. Article 50(2) of the EU AI Act requires providers to mark synthetic output so it is detectable as artificially generated, and California's act requires latent provenance data carrying the provider's name, system, time and a unique identifier. Neither law dictates a visible watermark on every image. Visible labels are required where a person must be told directly: EU deepfakes and public-interest AI text, chatbot notices, New York synthetic performer ads and political ad disclaimers. Platforms add their own label rules on top, which are contract terms rather than law.

AI use disclosure statements

A disclosure statement is a sentence or notice saying that AI was used and how. No US statute prescribes general wording, but the laws above set minimums: clear and conspicuous placement (California BOT Act, SB 243, Maine), timing at the start of a conversation and at intervals (New York), and the triggering request in Utah. In practice that means a short, plain notice in the chat window at the first message, repeated where New York's interval rule applies, rather than a line buried in terms of service. For publishers and universities, disclosure duties often come from editorial or academic policy rather than statute; see our compliance hub.

Outside the US and EU

South Korea's AI Basic Act, in force since January 22, 2026, requires generative AI operators to label AI output and to give clear notice when AI-generated sound, images or video are hard to tell from real ones, with fines of up to KRW 30 million after a grace period (Cooley).

Bottom line

As of November 9, 2026, there is no general US duty to label AI content, but specific laws require disclosure for chatbots that could pass as human, for AI image, video and audio from generative AI providers in California and the EU, for synthetic performers in New York ads and for AI in political ads in about 30 states. EU Article 50 has applied since August 2, 2026, with fines up to EUR 15 million or 3% of turnover.

Key facts at a glance. Sources listed below.

What changed

2026-11-12: Page published, with status of each disclosure law verified as of November 9, 2026.
2026-09-30: California SB 1000 took effect, removing the one million user threshold and the visible label option from the AI Transparency Act.
2026-08-02: EU AI Act Article 50 began to apply; California AI Transparency Act provider duties became operative.
2026-06-09: New York synthetic performer ad disclosure law (GBL § 396-b) took effect.
2026-01-22: South Korea AI Basic Act took effect, including AI output labeling duties.
2026-01-01: California SB 243 companion chatbot disclosure rules took effect.
2025-11-05: New York AI companion law (GBL Article 47) took effect.

Frequently asked questions

Do you have to disclose AI generated content?

Not as a general rule in the US. You must disclose in specific cases: chatbots that could be mistaken for people in several states, AI image, video and audio from generative AI providers in California, synthetic performers in New York ads, and AI-altered political ads in about 30 states. In the EU, Article 50 of the AI Act applies from August 2, 2026.

What is AI disclosure?

AI disclosure means telling people that AI produced content or is running an interaction. It can be a visible notice, such as a chatbot message or an ad label, or a machine-readable mark embedded in a file, such as the latent provenance data California requires in AI-generated images, video and audio.

Which laws require AI disclosure?

The main ones are EU AI Act Article 50, California's AI Transparency Act, BOT Act, SB 243 and, from 2027, AB 1609, New York's AI companion and synthetic performer laws, Utah's AI Policy Act, Maine's LD 1727, and about 30 state election deepfake laws. The FTC Act reaches deceptive undisclosed AI use.

Do chatbots have to say they are AI?

In the EU, yes, unless it is obvious, from August 2, 2026. In California, New York, Utah and Maine, chatbots must disclose in defined situations: when a person could be misled, at the start of companion conversations, or when a user asks directly. California adds a duty for large companies' customer service bots from January 1, 2027.

Sources

Legal disclaimer

This page is journalism and commentary, not legal advice. AI law changes quickly; check the primary sources above and consult qualified counsel for your situation.