US AI disclosure laws: federal, state and FTC rules for businesses

This page covers the US rules we have read in their primary texts. It is general information, not legal advice: whether a rule applies to you depends on facts about you, and several of these laws changed in the weeks before this review. For the EU rules, see EU AI Act Article 50; for platform rules, see AI disclosure rules by platform; for the whole map, start at the AI disclosure laws overview.

The rules at a glance

AI-generated people in advertising

Two laws ask an advertiser to disclose a synthetic performer: a digital figure, voice or representation that creates the impression of a human performer who is not recognizable as an identifiable real person. Both are about the people in the ad. A real, identifiable person's AI likeness is a consent question under other laws, and a label does not cure missing consent. Tennessee's ELVIS Act (Public Chapter 588, in force since July 1, 2024), for example, creates liability for using a person's voice or likeness in advertising without consent, and says nothing about disclosure.

New York: General Business Law 396-b

"3. Any person engaged in the business of dealing in any property or service who for any commercial purpose produces or creates an advertisement respecting any such property or service, in any medium or media in which such advertisement appears, shall conspicuously disclose in such advertisement that a synthetic performer is in such advertisement, where such person has actual knowledge. A violation of this subdivision shall result in a civil penalty of one thousand dollars for a first violation, and five thousand dollars for any subsequent violation."

The statute 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 and/or visual performance of a human performer who is not recognizable as any identifiable natural performer." Three details matter.

• Narrow trigger. The duty attaches only "where such person has actual knowledge". A seller that buys a stock avatar video without knowing it is synthetic is not caught by the text; one that generated it is.
• No set wording. The statute says "conspicuously" and nothing about wording, size, duration or placement.
• Exceptions. Advertisements for expressive works such as films, television, streaming, documentaries and video games, where the performer's use in the ad is consistent with its use in the work; audio advertisements; advertisements where AI only translates the language of a human performer; and media that publish or disseminate advertisements.

The act was signed on December 11, 2025 (Chapter 617) and takes effect on the 180th day after it became law, which we calculate as June 9, 2026. The statute names no enforcer, no cure period and no private right of action, so who may bring a case is not stated in the text. The sponsor memo describes a broader law requiring disclaimers for deepfakes in advertising and on social media; the enacted text is not that broad. We found no enforcement action.

Sources: New York General Business Law 396-b; Assembly bill A8887-B and action history. Verified October 21, 2026.

California: SB 1050, Business and Professions Code 17610

California's act was approved on September 16, 2026. It has no operative-date clause, so under the general rule it takes effect on January 1, 2027 (our derivation). Its core provisions:

"It is unlawful for any person to create and cause to be published in an advertising medium an advertisement that prominently includes a synthetic performer without a clear and conspicuous disclosure that the advertisement includes a synthetic performer."

"A disclosure required by this section shall use wording substantially similar to "this performance features a synthetic performer" or "no human performer is depicted.""

• Prominently. The duty applies only if the synthetic performer is in the foreground and demonstrating or illustrating the product or service; provides or voices the on- or off-camera narration or commercial message; or illustrates or reacts to that narration or message. A background extra is outside.
• Clear and conspicuous means "a disclosure that is difficult to miss, easily understandable, and presented in a manner that a reasonable consumer would notice, read, and comprehend, taking into account the medium, format, and context in which the advertisement appears."
• Audio is included. Unlike New York, California covers audio-only ads, and requires generative AI rather than "a software algorithm".
• Exceptions. Advertisements for expressive works, consistent with the work, and advertisements where generative AI solely involves language translation of a human performer or other accessibility features.
• Enforcement. "A violation of this section constitutes a violation of Section 17500 and may be enforced pursuant to Chapter 5 (commencing with Section 17200) of Part 2." That brings in the False Advertising Law and the Unfair Competition Law. An "advertising medium" such as a platform or publisher has a separate, narrower duty that arises only after a court order finding a violation.

Law firm summaries cite "up to $2,500 per violation"; the bill text states no amount, so we do not repeat it. For wording help, see our synthetic performer ads guide and the label generator. On Amazon, a related keyword in the image file is reported for photorealistic AI-generated people; see the platforms page. A tag in a file is not the label these laws describe: they are about what viewers of the ad see or hear.

Source: California SB 1050 (2026) bill text. Verified October 21, 2026.

Real estate: California AB 723

Business and Professions Code 10140.8 has applied since January 1, 2026 to "a real estate broker or salesperson, or person acting on their behalf", for images they create or have created in an advertisement or other promotional material for the sale of real property. It is not an AI-only law: the trigger is an image altered through photo editing software or AI to add, remove or change elements. A virtual staging vendor, photographer or marketing agency working for a licensee is plausibly a person acting on their behalf; a seller who is not a licensee is not bound by this section, and how it applies to an owner selling without an agent is untested.

A "digitally altered image" is:

"an image ... that has been altered through the use of photo editing software or artificial intelligence to add, remove, or change elements in the image, including, but not limited to, fixtures, furniture, appliances, flooring, walls, paint color, hardscape, landscape, facade, floor plans, and elements outside of, or visible from, the property, including, but not limited to, streetlights, utility poles, views through windows, and neighboring properties."

The duty is to include:

"a statement disclosing that the image has been altered and a link to a publicly accessible internet website, URL, or QR code that includes, and clearly identifies, the original, unaltered image. The statement shall be reasonably conspicuous and located on or adjacent to the image and shall include language indicating that the unaltered images can be accessed on the linked internet website, URL, or QR code."

If the posting is on a website the broker or salesperson controls, they may instead include the unaltered versions in the posting or link to a page that includes and clearly identifies them. Images where only "lighting, sharpening, white balance, color correction, angle, straightening, cropping, exposure, or other common photo editing adjustments are made that do not change the representation of the real property" are outside the duty.

The penalty comes from the Real Estate Law, not from AB 723 itself. Section 10185 makes a willful violation, or knowingly participating in one, a misdemeanor:

"guilty of a misdemeanor punishable by a fine not exceeding ten thousand dollars ($10,000), or by imprisonment in the county jail not exceeding six months, or by a fine and imprisonment."

Several vendor blogs give other figures; none match the statute. We found no Department of Real Estate bulletin or enforcement action on AB 723.

Outside California

We found no other state statute on altered listing photos. The National Association of Realtors' Code of Ethics, Article 12, asks members to "present a true picture" in their advertising, and Standard of Practice 12-10 extends that to internet content and images, but no Standard of Practice names AI or virtual staging. NorthstarMLS has published proposed guidelines, dated July 10, 2026, that say virtually staged or AI-enhanced photos "should be identified" and ask for an unaltered image of each staged room. The New York Department of State issued an alert on November 13, 2025 that creates no new rule and points to existing law requiring advertisements to depict a property accurately. In the EU, the Commission's icons page lists an AI-furnished empty apartment as a use case for its Partially AI-Modified icon; see EU deep fakes. Our virtual staging disclosure guide covers the practical side.

Sources: California Business and Professions Code 10140.8; California Business and Professions Code 10185; NAR Code of Ethics and Standards of Practice; NorthstarMLS proposed guidelines; New York Department of State alert. Verified October 21, 2026.

Chatbots

In the US, the chatbot disclosure duties we found sit on the business that uses the bot, not on the vendor that built it. That is the opposite of the EU's Article 50(1), where the duty sits on the provider. Any checklist for both has to say so.

California SB 1001

Business and Professions Code 17941, in force since July 1, 2019, applies to "any person" using a bot to communicate with another person in California online:

"It shall be unlawful for any person to use a bot to communicate or interact with another person in California online, with the intent to mislead the other person about its artificial identity for the purpose of knowingly deceiving the person about the content of the communication in order to incentivize a purchase or sale of goods or services in a commercial transaction or to influence a vote in an election. A person using a bot shall not be liable under this section if the person discloses that it is a bot."

The disclosure "shall be clear, conspicuous, and reasonably designed to inform persons with whom the bot communicates or interacts that it is a bot." The rule requires an intent to mislead and a sales or voting purpose, so a clearly labeled chatbot is the safe harbor. The chapter has no penalty section of its own.

California AB 1609

For businesses above the revenue threshold, B&P 22626 adds from January 1, 2027:

"(a) A large private business shall not represent that a customer service chatbot is a human. (b) A large private business shall provide a clear and conspicuous disclosure that the customer service chatbot is artificially generated and not human if a reasonable person interacting with the customer service chatbot is likely to be misled to believe that the person is interacting with a human."

The disclosure must be in the medium in which the chatbot interacts with the customer. Section 22627 adds a simple way to request a human agent and good-faith limits on waiting times. Public prosecutors enforce it, and "This chapter does not establish a private right of action."

Maine

"A person may not use an artificial intelligence chatbot or any other computer technology to engage in trade and commerce with a consumer in a manner that may mislead or deceive a reasonable consumer into believing that the consumer is engaging with a human being unless the consumer is notified in a clear and conspicuous manner that the consumer is not engaging with a human being."

It is the broadest of the three because it covers any person, of any size, and "any other computer technology". The act says "A violation of subsection 2 is a violation of the Maine Unfair Trade Practices Act." Maine's act does not print an effective date; Maine law generally takes effect 90 days after adjournment, which a law firm summary computes as September 23, 2025, and we did not verify that.

Utah

Section 13-75-103 requires a supplier using generative AI to interact with an individual in a consumer transaction to disclose that the individual "is interacting with generative artificial intelligence and not a human, if the individual asks or otherwise prompts the supplier about whether artificial intelligence is being used". The definition of generative AI is narrow: scripted bots are outside it. Section 13-75-104 offers a safe harbor:

"A person is not subject to an enforcement action for violating Section 13-75-103 if the person's generative artificial intelligence clearly and conspicuously discloses: (a) at the outset of any interaction with an individual in connection with: (i) a consumer transaction; or (ii) the provision of regulated services; and (b) throughout the interaction that it: (i) is generative artificial intelligence; (ii) is not human; or (iii) is an artificial intelligence assistant."

Violations are enforced by the Division of Consumer Protection, with administrative fines of up to $2,500 for each violation. Section 13-75-102 says it is not a defense that generative AI made the statement or took the act.

Where there is no duty

Texas's TRAIGA imposes no chatbot or AI-content disclosure duty on private businesses (see below). Colorado's 2026 replacement act contains no general "you are interacting with an AI" notice. A 2026 wave of companion-chatbot laws in Washington, Oregon, Idaho, Nebraska and Iowa asks for a "this is AI" notice when a reasonable person would be misled, but in the texts we read they exclude or partly exclude customer service and business-use bots; they take effect between January 1, 2027 and July 1, 2027 according to the texts and reports we read. For the practical side, see our AI chatbot disclosure guide.

Sources: California Business and Professions Code 17941; California AB 1609 (2026) bill text; Maine Public Law 2025, chapter 294 (PDF); Utah SB 226 (2025), enrolled (PDF). Companion chatbot laws: Washington HB 2225 (PDF), Oregon SB 1546, Idaho S1297 (PDF), Nebraska LB 525 (PDF) and Iowa SF 2417 (PDF). Verified October 21, 2026.

Provenance laws: providers and platforms, not sellers

A group of laws puts provenance duties on the companies that build generative AI systems and on large platforms. They shape what metadata a seller's images carry and what a platform will show, but they do not ask an ordinary shop to label anything.

California AI Transparency Act

Business and Professions Code section 22757.6 says "This chapter shall become operative on August 2, 2026." The same day, EU Article 50 began to apply. SB 1000, an urgency statute effective September 30, 2026, rewrote the chapter weeks after it started to apply. It binds:

• a "covered provider": "a person that creates, codes, or otherwise produces a generative artificial intelligence system that is publicly accessible within the geographic boundaries of the state". The current code text has no user threshold;
• a "large online platform": a social media, file-sharing, mass messaging or stand-alone search platform with more than 2,000,000 unique monthly users in the preceding 12 months, with duties from January 1, 2027; and
• generative AI hosting platforms, from January 1, 2027.

Covered providers must offer a free disclosure verification tool and, to the extent technically feasible, embed a latent disclosure in image, video and audio content they create or alter, naming the provider and system, the time and date and a unique identifier. Platforms must detect provenance data, show it and not knowingly strip it. The penalty is "$5,000 per violation to be collected in a civil action filed by the Attorney General, a city attorney, or a county counsel", and each day is a discrete violation. We found no enforcement action.

Other provenance laws bind providers above a user threshold: Connecticut Public Act 26-15, section 15 (more than one million users a month, from October 1, 2026), Utah HB 276 (over 1,000,000 monthly visitors or users, from January 1, 2027) and Washington HB 1170 (from February 1, 2027). A New York bill, the Stop Deepfakes Act, would add provenance duties for providers and large platforms; it has passed both houses and has not been delivered to the Governor.

Sources: California Business and Professions Code 22757.1; 22757.4; Utah HB 276 (2026), enrolled (PDF); Connecticut Public Act 26-15 (PDF). Verified October 21, 2026.

Notice duties for employers, AI subscriptions and consequential decisions

These laws are not labels on content. They are notices to the people an AI system affects. We include them because they bind ordinary businesses and they come up in the same searches.

• Illinois. Since January 1, 2026, 775 ILCS 5/2-102(L)(2) makes it a civil rights violation "For an employer to fail to provide notice to an employee that the employer is using artificial intelligence" for recruitment, hiring, promotion, discipline and similar decisions. The Illinois Department of Human Rights is to adopt rules on when and how notice is given. We did not verify that final rules exist.
• Connecticut. From October 1, 2026, section 1 of Public Act 26-15 binds a "subscription-based provider" of AI technology to Connecticut consumers. It may not enter into or renew a subscription, or collect a fee, unless it has given the consumer "a written notice disclosing the key terms and conditions of the subscription" and the consumer has accepted them in writing. The key terms include quantitative or qualitative limits and whether the provider can limit or eliminate access or reduce quality. The Attorney General enforces it, and there is no private right of action. It reaches any software company that sells AI functionality by subscription to Connecticut residents.
• Colorado. SB 26-189, signed May 14, 2026, repeals and replaces the Colorado AI Act and takes effect January 1, 2027. It keeps a notice regime for automated decision technology that materially influences consequential decisions in employment, education, housing, lending, insurance, health care and essential government services. Advertising, marketing, differentiated product recommendations, search and customer service triage are excluded. A separate 2026 Colorado chatbot law is reported by a secondary source, and we did not read it.

Sources: Illinois Public Act 103-0804; Connecticut Public Act 26-15 (PDF); Colorado SB 26-189, signed act (PDF). Verified October 21, 2026.

Texas

The Texas Responsible Artificial Intelligence Governance Act (HB 149) has applied since January 1, 2026. Its disclosure duties bind only a "governmental agency that makes available an artificial intelligence system intended to interact with consumers" and a health care provider that uses AI in a service or treatment (section 552.051). It sets no chatbot or AI-content disclosure duty on private businesses in general. Its other provisions prohibit certain uses of AI, and the Attorney General enforces it alone, after a 60-day written notice and cure period. We found no enforcement action.

Source: Texas HB 149 (2025), enrolled (PDF). Verified October 21, 2026.

The FTC and federal law

We found no US federal statute and no FTC rule that requires content to be labeled as AI-generated. The federal model is deception law: Section 5 of the FTC Act, the Consumer Reviews and Testimonials Rule and the Endorsement Guides. Four things from the FTC matter to a business that sells or uses AI tools.

• Claims about AI products need evidence. The final order against accessiBe (April 22, 2025, $1,000,000) bars claims, without evidence, that its automated product can make any website meet WCAG, and requires disclosure of paid reviews. The final order against Workado (August 28, 2025) requires "competent and reliable evidence" for accuracy claims about an AI content detector. This is why GotAlt does not claim to detect AI-generated images or to make a site meet any law.
• Fake reviews are banned however they are made. 16 CFR 465.2(a) makes it unfair or deceptive for a business to write, create or sell a consumer review or testimonial that "materially misrepresents, expressly or by implication: (1) That the reviewer or testimonialist exists", or that they used the product or what their experience was. The FTC said in announcing the rule that it covers AI-generated fake reviews, and its Q&A says the rule "has no blanket prohibition on the use of AI-generated avatars in marketing", while giving no safe harbor. On December 22, 2025 the FTC sent warning letters to ten unnamed companies about possible violations, citing civil penalties of up to $53,088 per violation. We found no FTC complaint that charges AI-written reviews as such.
• The Rytr order is gone. On December 22, 2025 the FTC set aside its order against Rytr, an AI writing tool that offered a review-generating use case, saying the order unduly burdened innovation. Part 465 is untouched, and the FTC said it would keep pursuing actors who use AI to violate the law.
• Endorsement Guides. The Guides (16 CFR Part 255) contain no AI wording. Their closest hook for AI-generated "customer" videos is the rule on advertisements that present endorsements by what are represented to be "actual consumers": use actual consumers, or clearly and conspicuously disclose that the people shown are not. That is our reading of the text, not an FTC statement about AI.

The TAKE IT DOWN Act is not a labeling law, but it is the federal rule that reaches AI-generated sexual deepfakes on user-generated-content sites. Platform duties began on May 19, 2026, when the FTC said it began enforcing the Act: covered platforms must remove nonconsensual intimate depictions, including AI "digital forgeries", within 48 hours of a valid request. The FTC sent letters to 15 named platforms. A Senate bill, S. 4915, the AI Labeling Act of 2026, was introduced on June 24, 2026 and referred to committee. It would require a visible label embedded in generative AI image, video and audio content that is accessible to people with disabilities where technically and economically feasible, plus machine-readable metadata. It is not law.

Sources: FTC, accessiBe final order; FTC, Workado final order; 16 CFR Part 465; FTC Q&A on the reviews rule; FTC warning letters, December 22, 2025; FTC, Rytr order set aside; 16 CFR Part 255; FTC, TAKE IT DOWN Act enforcement; S. 4915 bill text. Verified October 21, 2026.

Election deepfake laws and the court rulings

Most US state AI labeling law that has reached court is about political advertising. The Public Citizen legislative tracker, an advocacy group and so a secondary source, lists 33 states with an enacted election deepfake law as of October 21, 2026. They follow two models: a disclosure mandate for the sponsor of a political ad that uses AI-generated or manipulated media, and a prohibition on distributing deceptive deepfakes about candidates within a window before an election, with a label as a safe harbor or defense. The wording is fixed by statute and varies:

• California Government Code 84514(a)(1), for committees: "Ad generated or substantially altered using artificial intelligence."
• Florida Statutes 106.145(2): "Created in whole or in part with the use of generative artificial intelligence (AI)."
• Wisconsin Statutes 11.1303(2m), for audio: "Contains content generated by AI."
• Washington RCW 29B.35, as an affirmative defense: "This (image/video/audio) has been manipulated".

Courts have been less friendly to the stricter laws.

• Kohls v. Bonta (E.D. Cal.). California's AB 2839 was preliminarily enjoined on October 2, 2024, with the court calling its disclosure requirement compelled speech that is unduly burdensome. On August 29, 2025 the court permanently enjoined it as to the named plaintiffs. On August 20, 2025 the court permanently enjoined AB 2655, the platform law, as to X Corp. and Rumble, on Section 230 grounds, without reaching the First Amendment counts. California then rewrote AB 2839 by AB 502 (chaptered September 19, 2026); we found no ruling on the rewrite.
• Kohls v. Ellison (D. Minn., 8th Cir.). A challenge to Minnesota's criminal deepfake law lost a motion for a preliminary injunction, and on February 9, 2026 the Eighth Circuit affirmed without reaching the merits. The Minnesota law remains enforceable and its constitutionality is undecided. X Corp.'s separate case was dismissed in part on standing on December 2, 2025.
• Hawaii. The Public Citizen tracker marks Hawaii's SB 2687 as struck down. We did not read the ruling and treat that as unverified.

Small businesses are bound only if they run political or issue advertising, and marketing agencies that make political ads are the realistic audience. Our reading, not a legal conclusion: commercial speech mandates such as New York 396-b and California SB 1050 face a different and generally more forgiving standard than political speech mandates, and we found no challenge to either.

Sources: Public Citizen election deepfake tracker (secondary); Kohls v. Bonta records on govinfo.gov; Kohls v. Ellison records on govinfo.gov; Eighth Circuit opinion No. 25-1300 (PDF); Florida Statutes 106.145; Wisconsin Statutes 11.1303. Verified October 21, 2026.

Federal preemption: what the executive order does and does not do

Executive Order 14365 of December 11, 2025, "Ensuring a National Policy Framework for Artificial Intelligence", is an executive order, not a statute. It does not itself preempt any state law and creates no private rights. It directs the Attorney General to set up an AI Litigation Task Force "to challenge State AI laws inconsistent with the policy set forth in section 2", directs Commerce to identify "onerous" state laws, including laws that "may compel AI developers or deployers to disclose or report information in a manner that would violate the First Amendment", conditions some federal funding, and asks the FCC and FTC for further steps. Its legislative recommendation would not preempt state laws on child safety, state procurement or data center infrastructure.

What we could not verify is whether Commerce, the FTC or the FCC have taken each step, whether the Justice Department has sued any state over a disclosure law, or whether any court has ruled on the order. No state disclosure law on this page is verified as preempted or enjoined by federal action. In our reading, the order's stated targets point at developer-facing laws rather than at advertising disclosures and chatbot notices, but that is an inference, not a legal conclusion. Treat each rule's status as something to re-check on its date.

Source: Executive Order 14365, whitehouse.gov; 90 Federal Register 58499. Verified October 21, 2026.

Passed or proposed, not law yet

The New York legislature adjourned on June 5, 2026. Several bills passed both houses, but the Assembly's action histories show none delivered to the Governor, who has until December 31, 2026 to act once a bill is delivered.

• S.8451-B, the FAIR News Act. It would require "news media", defined broadly enough to include websites and newsletters, to label content substantially created by generative AI on the page. A small business blog or newsletter that publishes AI-drafted posts could arguably be caught by the literal definition. It has no human-review exception, but it exempts content eligible for copyright registration.
• S.6954-B, the Stop Deepfakes Act. Provenance duties for synthetic content providers and large online platforms.
• A.3411-B. A conspicuous notice that outputs of a generative AI system "may be inaccurate".

Treat these as watch items. Sources: S.8451-B and its action history; A.3411-B and its action history. Verified October 21, 2026.

Key dates