United States: AI content labelling rules

The rules on labelling AI-generated content in the United States, what each asks of a file, and what one Kennmarq.com call adds.

In short

Images: Covered
Video: Covered
Audio: Covered
Text: Not required
3D: Not required
Rules: 4 general, 30 for some sectors only, 4 bills or drafts.
In the API: markets=us

Updated 4 November 2026; sources checked 4 November 2026. Sources are government sites. This page explains what Kennmarq.com does; it isn't legal advice.

Political-ad AI rules exist in more states than listed; we list those whose text we could check. Texas, Minnesota and Maryland ban election deep fakes.

General rules

AI Transparency Act (SB 942, SB 1000) (California): In force since 2 August 2026. Latent disclosure: provider, system and version, time, unique ID, created or altered; free verification tool.
Public Act 26-15, section 15 (Connecticut): In force since 1 October 2026. Provenance data showing the provider's AI made or materially altered the content, hard to remove.
Digital Content Provenance Standards Act (HB 276) (Utah): Passed; applies from 1 January 2027. Latent disclosure with time and date and digital signatures proving AI creation, per industry standards.
AI content provenance law (HB 1170) (Washington): Passed; applies from 1 February 2027. Provenance data showing the provider's AI made or materially altered the content, hard to remove.

Rules for specific sectors

Political ads and campaigns

Alabama HB 172 (2024) (Alabama): In force since 1 October 2024. A disclaimer that the media was manipulated by technical means and shows speech or conduct that did not occur.
A.R.S. § 16-1023 (HB 2394, 2024) (Arizona): In force since 21 May 2024. Convey that it is a digital impersonation, or the person shown can get a court declaration against it.
A.R.S. § 16-1024 (SB 1359, 2024) (Arizona): In force since 14 September 2024. A clear and conspicuous disclosure that the media includes content generated by AI.
Cal. Government Code § 84514 (AB 2355, 2024) (California): In force since 1 January 2025. 'Ad generated or substantially altered using artificial intelligence.', shown like the ad's other disclosures.
Cal. Elections Code § 20010 (AB 730, 2019) (California): In force since 1 January 2020. 'This image/video/audio has been manipulated.', as large as the largest other text, for the whole video.
C.R.S. § 1-46-103 (HB 24-1147, 2024) (Colorado): In force since 1 July 2024. A set statement that it was edited and falsely appears authentic, also in metadata with tool and time.
15 Del. C. § 5145 (HS 1 for HB 316, 2024) (Delaware): In force since 9 October 2024. 'This image/video/audio has been altered or artificially generated.', as large as the largest other text.
Fla. Stat. § 106.145 (HB 919, 2024) (Florida): In force since 1 July 2024. The set disclaimer on political ads whose AI content shows a real person doing what never happened.
Indiana Code 3-9-8 (HEA 1133, 2024) (Indiana): In force since 12 March 2024. 'Elements of this media have been digitally altered or artificially generated.', for the whole video.
KRS 117.322 (SB 4, 2025 Ky. Acts ch. 66) (Kentucky): In force since 24 March 2025. Clear, conspicuous notice on deepfake audio or video of a candidate in the last 45 days; a court can order it.
La. R.S. 18:1463(F) (HB 459, Act 806 of 2026) (Louisiana): In force since 1 August 2026. A clear disclosure that it was created using AI, in the form of the paid-for-by notice.
21-A M.R.S. § 1014(5-C) (LD 517, 2026) (Maine): In force since 29 July 2026. Set words on ads with deceptive media of a candidate; the Ethics Commission sets size and placement.
Mississippi SB 2577 (2024), Laws 2024 ch. 467 (Mississippi): In force since 1 July 2024. Clear, prominent words displayed throughout that the person did not say or do it: a defence (audio has none).
MCA § 13-35-802 (SB 25, 2025) (Montana): In force since 1 October 2025. A statement that AI significantly edited it and it falsely appears authentic, within 60 days of voting.
NDCC § 16.1-10-04.2 (HB 1167, 2025) (North Dakota): In force since 1 August 2025. 'THIS CONTENT GENERATED BY ARTIFICIAL INTELLIGENCE.' stated prominently on AI that impersonates a person.
N.M. Campaign Reporting Act AI rules (HB 182, 2024) (New Mexico): In force since 15 May 2024. A disclaimer that AI manipulated or generated the media, on ads that misrepresent a person.
N.Y. Election Law § 14-106(5) (2024) (New York): In force since 20 April 2024. 'This image/video/audio has been manipulated.' in the communication's language, no smaller than its text.
ORS 260.268 (SB 1571, 2024) (Oregon): In force since 27 March 2024. A disclosure that the campaign communication uses synthetic (AI-manipulated) media.
SDCL §§ 12-26-32 to 12-26-37 (SB 164, 2025) (South Dakota): In force since 1 July 2025. 'This image/video/audio has been manipulated or generated by artificial intelligence.' over each deepfake.
T.C.A. § 2-19-120(a)(6) (HB 1513, Pub. Ch. 625, 2026) (Tennessee): In force since 1 July 2026. A set statement in large letters for the whole video, or spoken at the start and end of audio.
Utah Code § 20A-11-1104 (SB 131, 2024) (Utah): In force since 1 May 2024. Set AI wording throughout AI parts, spoken at start and end of AI audio; provenance on online ads.
17 V.S.A. § 2032 (S.23, Act 75 of 2026) (Vermont): In force since 5 March 2026. A statement that it was digitally manipulated or generated and shows speech or conduct that did not occur.
RCW 29B.35 (SB 5152, 2023) (Washington): In force since 23 July 2023. 'This image/video/audio has been manipulated.', as large as the largest other text, for the whole video.

Advertising and influencers

California altered photos in rental ads (AB 2025) (California): Passed; applies from 1 January 2027. Notice on or next to an AI-made or altered photo that it was altered, plus the original or a link to it.
California altered photos in property sale ads (AB 723) (California): In force since 1 January 2026. Statement on or next to an altered photo that it was altered, with a link or QR code to the original.
California synthetic performers in ads (SB 1050) (California): Passed; applies from 1 January 2027. Clear, conspicuous notice that the ad features a synthetic performer.
New York synthetic performers in ads (GBL 396-b) (New York): In force since 9 June 2026. Conspicuous notice in the ad that a synthetic performer appears in it.
Pennsylvania AI Deceptive Advertising Act (Act 99 of 2026) (Pennsylvania): Passed; applies from 21 December 2026. Notice that the ad contains AI-generated or AI-modified content, from its first appearance, hard to remove.

Platforms

California AI Transparency Act: platforms (AB 853, AB 2713) (California): Passed; applies from 1 January 2027. Keep provenance data and signatures; show users if content is AI-made or camera-made and let them inspect it.
Utah HB 276: platform and capture-device provenance duties (Utah): Passed; applies from 1 January 2027. Detect and keep compliant provenance data and signatures; show users it exists and let them inspect it.

Bills and drafts

Not law: they don't change a country's coverage or what a call needs.

FCC proposed rule on AI in political ads (FCC 24-74): Draft: not in force. An on-air notice, before or during the ad, that it contains AI-generated content.
Stop Deepfakes Act (S6954B) (New York): Draft: not in force. Content credential: AI-made or edited, provider, time, AI-made parts, system type; platforms show it.
New York FAIR News Act (S8451B / A8962B) (New York): Draft: not in force. Notice that the content was substantially created by generative AI, imprinted or spoken at the start.
Pennsylvania HB 811 (2025) (Pennsylvania): Draft: not in force. A deepfake statement each time the ad shows, plus tamper-evident provenance, within 90 days of a vote.

In the API

Send markets=us on each call to POST /v1/mark: a call that lacks what the United States' rules need is refused before anything is counted, and each mark gets a receipt of what it carries for each law. The call must also send provider, app_name and app_version.

What a mark for the United States carries, by kind of file
File	Made with AI	A deep fake (`deepfake=true`)
Picture	Watermark and Content Credentials	Watermark and Content Credentials
Clip	Watermark and Content Credentials	Watermark and Content Credentials
Song	Watermark and Content Credentials	Refused for now: a song can't carry a visible label, and spoken notices for deep fakes aren't available yet: leave "deepfake" off for songs.

Mark a picture for the United States

curl https://api.kennmarq.com/v1/mark \
  -H "Authorization: Bearer $KENNMARQ_API_KEY" \
  -F file=@image.png \
  -F source=created \
  -F provider="Acme Inc." \
  -F app_name="Acme Studio" -F app_version="2.1" \
  -F markets=us \
  -o image.marked.png

An AI coding agent gets the same answer from the MCP tool get_requirements (MCP server).

The United States on the map, and every country's rules side by side.