Do you have to disclose AI? The rules by platform and by law

Meta's AI-disclosure tool and "AI info" labels (Community Standards, Misinformation: Manipulated Media; Help Center, How to identify AI content on Meta products)

Yes, for realistic AI video and audio. Not required for images, though Meta may label them anyway.

Who must disclose

Anyone posting organic (unpaid) content on Facebook, Instagram or Threads, including an agency posting for a client. Meta's ads have separate rules.

What triggers it

Photorealistic video or realistic-sounding audio that was digitally created, modified or altered, including with AI (Meta's examples: a realistic video of people at an outdoor market, two people talking in an audio file, a song with AI-generated vocals, a reel narrated with a realistic AI voiceover). Meta says the requirement does not apply to images, but an image can still be labeled if Meta's systems detect that it was AI-generated or modified with AI.

How

The poster turns on Meta's AI label when sharing (on Threads: create a post and choose Add AI label), and the post then shows the "AI info" label. Separately, Meta reads industry-standard AI signals in uploaded files (the C2PA and IPTC "AI generated" metadata) and labels content itself: content with signals showing it was created with AI gets the AI info label on the post, while content with signals showing it was only modified with AI gets the information in the post's menu instead of on the post. Content made with Meta's own AI tools may be labeled automatically or carry a visible watermark.

Not required

Images, even when made or edited with AI (Meta does not require the creator label for images), and clearly stylized video such as a cartoon-style landscape, per Meta's own examples.

If you don't

Meta says there may be penalties for content shared without a label when one is required, and does not list them. Meta also adds the AI info label itself when it detects AI signals, and may add a more prominent label (or reject an ad) when digitally created or altered content creates a particularly high risk of materially deceiving the public on a matter of public importance.

For an agency

When you publish a client's realistic-looking video or lifelike AI voice on Facebook, Instagram or Threads, switch on the AI label yourself, because Meta may penalize a missing label and adds one anyway when the file carries AI metadata.

Sources: How to identify AI content on Meta products; Misinformation (Community Standards, Manipulated Media); Labeling AI-Generated Images on Facebook, Instagram and Threads (2024-02-06); Our Approach to Labeling AI-Generated Content and Manipulated Media (2024-04-05); Label AI content on Threads

AI-generated profile label (renamed from the "AI creator" label), Instagram Creators blog, 31 August 2026

Yes, when the account features an AI-generated person instead of a human.

Who must disclose

The owner of an Instagram profile that features an AI-generated person instead of a human (a virtual influencer or AI persona account).

What triggers it

A profile that features an AI-generated person rather than a human. Instagram says people dislike finding out later that a profile that seemed human features an AI-generated person.

How

The creator edits the profile and toggles on "AI-generated profile". The label then shows on the profile and alongside content whose media does not already carry an AI label. Creators who used the old "AI creator" label are asked to review it, confirm it still fits or remove it.

Not required

Creators who simply use AI tools as part of their creative process do not need to add the label.

If you don't

Unlabeled profiles that Instagram detects may have their reach limited and can become ineligible for recommendations. Instagram notifies them, they can check Account Status, and they can regain eligibility by adding the label or by appealing. Creators who add the label proactively will not see a change in reach.

For an agency

If you run an AI-generated persona account for a client on Instagram, turn on the AI-generated profile label or the account can lose recommendations; ordinary use of AI tools behind a human-fronted account does not trigger it.

Sources: Bringing More Transparency to AI-Generated Profiles on Instagram (2026-08-31); How to identify AI content on Meta products

Community Guidelines, Integrity and Authenticity: Edited Media and AI-Generated Content (AIGC); Help Center: About AI-generated content

Yes, for AI-generated or significantly edited content that looks real.

Who must disclose

Any TikTok creator, including brand and agency accounts.

What triggers it

AI-generated or significantly edited content that shows realistic-looking scenes or people. The Help Center says TikTok requires a label on all AI-generated content with realistic images, audio or video, and encourages one on content completely generated or significantly edited by AI (for example a real person made to say or do something they did not, or an AI face-swap).

How

Turn on the "AI-generated content" setting in the post settings before publishing (it adds a "creator labeled as AI-generated" label), or say so in text, a hashtag sticker or the description. Posts sent through TikTok's Content Posting API can set is_aigc to true for the same label. TikTok also applies an "AI-generated" label itself to content made with TikTok AI effects and to uploads carrying C2PA Content Credentials, and a creator cannot remove an automatic label. TikTok says turning the setting on does not affect distribution as long as the content follows the Community Guidelines.

Not required

Content made only with TikTok's own AI effects (TikTok labels it automatically), and minor corrections or enhancements, which the Help Center does not count as "significantly edited".

If you don't

Unlabeled content may be removed, restricted, or labeled by TikTok's team, depending on the harm it could cause. Putting the label on unaltered content breaks TikTok's Terms of Service and may get that content removed. Some AI content is not allowed even when labeled: fake authoritative sources or crisis events, public figures falsely shown being bullied or endorsing something, and the likeness of under-18s or of private adults used without permission.

For an agency

Label any client video or image that looks real and was generated or heavily edited with AI, using the in-app setting or the is_aigc field in your scheduling tool, and do not use AI likenesses of real people without their permission.

Sources: Integrity and Authenticity (Community Guidelines) (2026-09-24); About AI-generated content; Content Posting API: Direct Post reference; Integrity and Authenticity (Community Guidelines, version effective 13 September 2025) (2025-09-13)

Disclosing use of GenAI content (YouTube Help; launched in March 2024 as the altered or synthetic content disclosure)

Yes, for realistic AI content and for AI music that is the focus of the video.

Who must disclose

The creator or channel owner uploading the video or Short (the disclosure is made in YouTube Studio during upload).

What triggers it

AI-generated or meaningfully AI-altered content that seems realistic: content that makes a real person appear to say or do something they did not do, alters footage of a real event or place, generates a realistic scene that did not occur, or creates music that is the main focus of the video. It applies to content that is fully or partly altered or created with AI tools.

How

In YouTube Studio, on a computer or mobile device, answer Yes under "AI use" in the Attributes section during upload. YouTube then labels the video: in the video player for AI music or photorealistic content, or in the expanded description for non-music AI audio and non-photorealistic or animated content. YouTube may also apply a label itself for content made with its own GenAI tools, content containing C2PA metadata, and content its systems detect as AI-generated or altered; a creator cannot change a label from YouTube's own tools, from C2PA metadata or from manual review.

Not required

Content that is not realistic (for example someone riding a unicorn, or an animated missile in a fully animated video) and minor edits that are primarily aesthetic and do not mislead about what happened: beauty filters, color or lighting adjustments, background blur, using AI to create or improve a script, outline, thumbnail, title or infographic, captions, upscaling or sharpening, voice or audio repair, idea generation, cloning one's own voice for voice-overs or dubs, video game footage, and AI-generated or extended backdrops such as a moving car.

If you don't

Creators who consistently do not disclose may get a label applied by YouTube or face penalties, including removal of content or suspension from the YouTube Partner Program. Disclosing alone does not limit a video's audience or its eligibility to earn money. YouTube also applies its Community Guidelines to AI-generated content.

For an agency

Set "AI use" to Yes on every client upload that contains realistic AI footage, an AI-generated realistic scene or AI music, and keep a note of which uploads are exempt (AI scripts, thumbnails and cloned own-voice voice-overs do not count).

Sources: Disclosing use of GenAI content; How we're helping creators disclose altered or synthetic content (2024-03-18)

Professional Community Policies (do not share false or misleading content, synthetic or manipulated media) and Content Credentials (C2PA) label

Only for fake or altered media of a real person. Otherwise LinkedIn recommends, but does not require, saying so.

Who must disclose

Any member or Page posting on LinkedIn. There is no general duty to disclose AI use; the duty applies to synthetic or manipulated media of real people.

What triggers it

Synthetic or manipulated media that depicts a person saying something they did not say or doing something they did not do must clearly disclose its fake or altered nature. For other AI-assisted content LinkedIn only recommends telling readers if you relied heavily on AI.

How

The policy asks for the fake or altered nature to be clearly disclosed and names no mechanism; LinkedIn describes no disclosure toggle. Separately, image and video carrying signed C2PA Content Credentials gets a C2PA icon that any viewer can click to see whether AI was used, the app or device, the creator and the issuer. The icon comes from the file's metadata, not from the poster. On Sponsored Content it shows only on single image or video ads.

Not required

AI-assisted writing, editing or images that do not depict a real person doing or saying something they did not. LinkedIn says AI-assisted content is welcome when it reflects a real person's perspective, and that it focuses on whether content adds value, not how it was made.

If you don't

Depending on severity LinkedIn may limit the visibility of content, label it or remove it, and repeated or egregious offenses can lead to account restriction. LinkedIn also says generic, low-value AI content ("AI slop") is less likely to be widely distributed.

For an agency

LinkedIn has no AI toggle for posts, so write a plain disclosure into any client post that shows a real person in AI-made or altered media, and expect a C2PA icon if your tool embeds Content Credentials in the file.

Sources: Professional Community Policies; Content credentials; Best practices for content created with the help of AI; LinkedIn adopts C2PA standard (2024-05-15)

X Authenticity policy, Inauthentic Content: Synthetic and Manipulated Media

No disclosure duty. Misleading AI media can be restricted or removed.

Who must disclose

Anyone posting on X. The policy bans misleading media; it does not set a general duty for creators to disclose AI use.

What triggers it

Inauthentic media, including manipulated or out-of-context media, that may result in widespread confusion on public issues, impact public safety, or cause serious harm ("misleading media"). This includes media shared as authentic that was significantly and deceptively altered, and media depicting a real person that was fabricated or simulated, especially with AI.

How

No creator disclosure step is defined in this policy. X uses its own technology and third-party reports to judge whether media was manipulated or out of context, and says it may not act when it cannot reliably tell. X also says it has expanded labeling of manipulated and AI-generated media, including "Made with AI" indicators, for users in the EU.

Not required

Sharing manipulated or out-of-context media in non-deceptive ways is allowed.

If you don't

Depending on severity and history X can restrict a post's reach, temporarily limit features, require deletion of posts, lock the account, or suspend it (permanently at first detection for severe violations).

For an agency

X does not ask you to tag AI content, but AI media that fabricates a real person or is presented as real footage can be restricted or suspended, so avoid that and say plainly in the post when an image or video is AI-made.

Sources: Authenticity (2025-04); X Media Literacy Action Plan (2026-07)

Gen AI labels ("AI modified") and the "Mark as AI-Modified" setting when creating a Pin

No duty. Pinterest labels AI Pins it detects, and you can mark them yourself.

Who must disclose

Pinterest applies the label automatically, and the Pin's creator can also mark a Pin as AI-modified. Pinterest presents the creator setting as an option, not a requirement.

What triggers it

Pins that were generated or modified using AI, and Pins that include an AI-generated person.

How

Pinterest adds an "AI modified" label, shown at the bottom left of a Pin in close-up, when it detects AI use from the Pin's IPTC metadata or through its classifiers. A creator can also use the "Mark as AI-Modified" toggles in the Pin creation form to say the content was made completely or partly with AI or includes an AI-generated person. For ads, the disclosure sits behind the ellipsis menu next to the Pin instead.

If you don't

Pinterest adds the label itself when it detects AI, and a creator who thinks a label is wrong can appeal through Pinterest support. Pinterest names no penalty for not marking a Pin, and its Community Guidelines apply to synthetically generated content like any other.

For an agency

Pinterest states no creator duty to mark AI content, but it labels such Pins itself from metadata and classifiers, so use the toggles on client Pins made with AI to match what Pinterest will show.

Sources: Gen AI labels; Create a Pin from an image or video; Introducing Gen AI Labels: Pinterest is Taking a New Step in Transparency (2025-04-30); Community Guidelines (current) (2026-05)

Pinterest Community Guidelines, Scams, deceptive goods, and practices (version effective 12 November 2026)

Not a label rule: from November 12, 2026, AI media that misrepresents a product is banned in posts and ads.

Who must disclose

Anyone posting on Pinterest, including advertisers and merchants.

What triggers it

Synthetically generated or manipulated media used to misrepresent the appearance, efficacy, or nature of a product. This sentence appears in the preview of the updated guidelines; a text search of the version in force today (last updated May 2026) did not find it.

How

This is a prohibition, not a disclosure step: no label or toggle is described for it.

If you don't

Pinterest says it may remove, limit or block distribution of violating content and the accounts or domains behind it, based on how much harm it poses.

For an agency

From 12 November 2026, do not use AI-generated or edited images or video that show a client's product looking or working differently from the real product on Pinterest.

Sources: Community Guidelines (preview of the version effective 12 November 2026) (2026-11); Community Guidelines (current) (2026-05); Advertising Guidelines (updated version, preview) (2026-11-12); Advertising guidelines (current version)

Community Guidelines (generative AI), Content Guidelines for Recommendation Eligibility (Spotlight), Creator Monetization Policy

Only where local law requires it, and on monetized content.

Who must disclose

Snapchatters who post AI-generated or AI-edited content, and creators monetizing content.

What triggers it

Snap sets no single disclosure toggle for users. Its Community Guidelines say some regions require users to disclose or label AI-generated content and that the user is responsible for meeting them. Separately, wholly AI-generated videos are no longer eligible for Spotlight recommendation, and monetized content that uses AI must disclose the AI use.

How

For monetization, AI use must be disclosed anywhere in the content or in the creator profile. Content made or edited with Snapchat's own AI tools carries Snap's transparency indicators automatically. Snap describes no AI label control for creators.

Not required

Content enhanced or edited with Snapchat's AI creative tools stays eligible for recommendation, and Snap says its tools always include transparency indicators.

If you don't

Wholly AI-generated video is not eligible for Spotlight recommendation, even with a disclosure (it can still be posted). Snap may take enforcement action, up to account termination, against accounts that use AI to break its Community Guidelines, abuse its AI products or break local laws. Monetization requires that AI use is disclosed.

For an agency

Do not count on Spotlight reach for fully AI-generated client video, disclose AI use in the content or profile of any monetized account, and follow the label rules of the region you publish in.

Sources: Community Guidelines; Rewarding Authentic Creativity on Spotlight (2026-07-31); Content Guidelines for Recommendation Eligibility; Creator Monetization Policy

Google Ads Political content policy, section 'Disclosure requirements for synthetic or digitally altered content'

Yes, for election ads with realistic synthetic people, voices or events.

Who must disclose

Verified election advertisers, in the regions where Google requires election ads verification, who run election ads on Google's ad products.

What triggers it

Any election ad that contains synthetic or digitally altered images, video or audio that inauthentically depict real or realistic-looking people or events. Google's two examples: an ad that makes it appear a person said or did something they did not, and an ad that alters footage of a real event or generates a realistic portrayal of an event that did not take place.

How

The advertiser ticks the checkbox 'Altered or synthetic content' in the campaign settings. For feeds on mobile phones, Shorts on mobile phones and in-stream formats, Google then generates the in-ad disclosure itself. For every other format the advertiser must also put a clear and conspicuous disclosure in the ad, in a place users are likely to notice. Google's example wordings: 'Altered or synthetic content.', 'This audio was computer generated.', 'This image does not depict real events.', 'This video content was synthetically generated.'

Not required

Synthetic or digital edits that are inconsequential to the claims made in the ad: resizing, cropping, color or brightening corrections, defect correction such as red-eye removal, and background edits that do not create realistic depictions of actual events.

If you don't

Google says a violation of this policy will not lead to immediate account suspension without prior warning, and that it will issue a warning at least 7 days before suspending the account.

For an agency

If you run election ads for a client where Google requires election ads verification, tick 'Altered or synthetic content' on every campaign with AI-made or AI-altered realistic people, voices or events. Other ads fall under Google's separate AI label setting.

Sources: Political content; Update to our policy on Disclosure requirements for synthetic content (July 2024) (2024-07-01); Updates to Political content policy (September 2023) (2023-09)

Google Ads policy change 'Updates to AI labeling requirements (July 2026)' and help article 'Use AI content label settings and disclosures'

Where the law requires it (the EU, India and New York). Since July 2026 any ad can be labeled.

Who must disclose

Advertisers whose ad images or videos were generated or modified with AI, in Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center and Google Ads Editor. Google also labels assets made with its own AI tools.

What triggers it

Ad assets created or edited with AI. Google ties labeling to laws in the EU, India and New York that require certain AI-generated or edited ad assets to carry a disclosure. Elsewhere Google offers the label as a setting, not as a duty of its own.

How

In the AI label setting, an advertiser chooses 'Label this asset as created or edited with AI' or 'Don't label this asset' when reviewing assets in a campaign or in the Asset library; or the advertiser puts a visible label inside the image or video itself. A labeled ad gets an entry in the 'How this ad was made' section of the My Ad Center panel (three-dot menu on ads across Search, YouTube and Discover). For campaigns targeting the EU, India or New York, ads with assets marked AI created or edited also show a visible overlay on the ad. Google may label assets itself (for example fully automated creatives or regions where Google must) and those labels cannot be overwritten. Uploading a video to YouTube from inside Google Ads prompts a question on whether AI was used, for YouTube's disclosure policy on organic videos.

If you don't

Google names no penalty for leaving an asset unlabeled, and says using the setting does not by itself guarantee compliance with any law. Misleading and deceptive ads stay prohibited, made with AI or not.

For an agency

Mark AI-made or AI-edited client creatives with the AI label setting (or label them inside the asset) whenever a campaign targets the EU, India or New York, and get legal advice because Google says the setting does not by itself prove compliance.

Sources: Updates to AI labeling requirements (July 2026) (2026-07-09); Use AI content label settings and disclosures; Expanding AI transparency in ads (2026-07-09)

Meta Business Help Center, 'About media created or edited with AI for ads about social issues, elections, or politics'

Yes, for social issue, election and political ads with realistic AI media.

Who must disclose

Advertisers running ads about social issues, elections or politics on Facebook and Instagram, set up in the Special Ad Category 'Social issues, elections or politics'.

What triggers it

A photorealistic image, video or realistic-sounding audio created or edited with AI tools (third-party tools included) that depicts a real person saying or doing something they did not; depicts a realistic-looking person who does not exist or a realistic-looking event that did not happen, or alters footage of a real event; or depicts a realistic event that allegedly occurred but is not a true recording of it.

How

In Meta Ads Manager the advertiser selects 'Social issues, elections or politics' in the Special Ad Category section, then selects 'Includes media created or edited with AI' in the Ad creative section. Meta then shows an 'AI info' label near the 'Paid for by' disclaimer, and the Ad Library entry says 'Media in this ad created or edited with AI.' Meta also runs automated detection of media made with third-party AI tools, which needs no advertiser action.

Not required

AI edits that are inconsequential or immaterial to the claim, assertion or issue raised in the ad, for example image size adjusting, cropping, color correction and image sharpening.

If you don't

The ad may be removed, and repeated failure to disclose may result in penalties on the account.

For an agency

For a client's social issue, election or political ads on Meta, tick 'Includes media created or edited with AI' whenever a realistic image, video or voice was made or altered with any AI tool, because Meta can also detect AI media itself.

Sources: About media created or edited with AI for ads about social issues, elections, or politics; How Meta Is Planning for Elections in 2024 (2023-11-28); About AI info on ads created or edited with generative AI tools

Meta Business Help Center, 'About AI info on ads created or edited with generative AI tools'

No filing needed. Meta labels ads made with its own AI tools or detected AI tools.

Who must disclose

Meta applies the label to Facebook and Instagram ads made or edited with its own generative AI tools, and to ads it detects were made or edited with third-party AI tools. In some regions the advertiser can also self-disclose.

What triggers it

Ad images or videos created or significantly edited with Meta's generative AI features (Background Generation, Image Generation, Add Animation), and ads Meta detects as created or edited with third-party generative AI tools such as Photoshop or DALL-E.

How

The label is applied automatically and found through the three-dot menu, then 'About this ad', then 'AI info'. Meta detects third-party AI use through industry-standard signals such as C2PA metadata. Before publishing, Ads Manager shows a message when an AI info label will appear on the ad. For ads delivered to the European Region, California, New York, India and Taiwan, advertisers may get the chance to self-disclose that GenAI was used, and the label then appears on the ad next to the Sponsored label (a photorealistic AI-generated human made with Meta's tools also puts the label next to Sponsored).

Not required

Meta does not label minor or immaterial changes made with its generative AI tools, such as image resizing or color correction, nor edits that are not significant and include no photorealistic human.

If you don't

Meta adds the label itself and names no penalty around it. Ads about social issues, elections or politics carry their own disclosure duty.

For an agency

Ordinary client ads need no Meta filing for AI use, but Meta will label AI-made creative on its own, including creative from outside tools that carries C2PA metadata, so tell clients before launch; use the self-disclosure option where Meta offers it in the EU, California, New York, India and Taiwan.

Sources: About AI info on ads created or edited with generative AI tools; Expanding GenAI Transparency for Meta's Ads Products (2026-06-01); How AI-generated images in ads are identified and labelled on Meta

TikTok Advertising Policies, 'Misleading and false content' (Edited media and AI-generated content) and Ads Manager help 'About ad disclaimers' / 'How to add disclaimers to ads'

Yes, for ads that are fully AI-generated or significantly edited with AI.

Who must disclose

The advertiser, for TikTok ads that are not Spark Ads. Spark Ads run as organic posts and follow TikTok's Community Guidelines on AI-generated content disclosure instead.

What triggers it

Image, video or audio that is completely AI-generated, or real source material significantly modified by AI beyond minor tweaks. TikTok's examples: showing the primary subject doing something they did not do, or making them say something they did not say using AI voice-cloning.

How

In TikTok Ads Manager, at ad creation, check the toggle 'This ad contains AI-generated content'; it adds a text disclosure label to the ad. TikTok's ad policy also accepts the AIGC label or the advertiser's own clear disclaimer, caption, watermark or sticker. After submission the toggle cannot be changed unless the video or image is replaced, and a new or duplicated campaign resets it.

Not required

Insignificant AI edits: adjustments to lighting, brightness or color saturation, removing or modifying backgrounds, and denoising. TikTok adds that if you are unsure, you should label.

If you don't

If TikTok identifies AI-generated content that was not disclosed, the ad is rejected or restricted. Using the disclaimer feature for anything other than disclaimer information also gets the ad rejected.

For an agency

Switch on 'This ad contains AI-generated content' in Ads Manager for any TikTok ad with fully AI-made or heavily AI-edited visuals or voices, and re-check it on every duplicated campaign because duplication resets the toggle.

Sources: About ad disclaimers in TikTok Ads Manager (2025-09); How to add disclaimers to ads in TikTok Ads Manager (2025-09); Misleading and false content (2026-04)

Microsoft Advertising help, 'AI-generated and synthetic content in advertising'

Wherever the law requires it. Deepfakes are banned even when disclosed.

Who must disclose

The advertiser, for any ad that uses AI-generated, AI-manipulated or other synthetic content.

What triggers it

Content generated or materially changed with AI. Microsoft does not set its own universal label duty: the advertiser must follow the laws that apply where the ad is shown, which includes disclosures when required, and must keep watermarks, metadata and other provenance information that shows how the content was created. Some content, such as deepfake content, is prohibited even if disclosed.

How

Advertisers should embed AI disclosures directly into image and video assets, close to the relevant content and clear to understand, and may use the 'Disclaimer in ads' feature where the ad format supports it. Microsoft's own AI tools add machine-readable provenance information, metadata and imperceptible watermarks, which may not show as a visible label.

If you don't

Microsoft Advertising may reject, restrict or remove ads that contain prohibited deepfake content or otherwise use AI to mislead, impersonate, omit a required disclosure, use a person's likeness or voice without authorization, or strip required machine-readable information. A disclosure does not make deceptive content acceptable.

For an agency

On Microsoft Advertising, add any legally required AI disclosure to the creative itself, keep the AI provenance metadata in the files you upload, and do not use deepfakes of real people; Microsoft also does not allow political advertising at all.

Sources: AI-generated and synthetic content in advertising; API interfaces for political ads policy enforcement (2025-10-02); Microsoft Advertising Publishes AI Content Transparency Rules (2026-09-14)

Snap Advertising Policies, section 4.3 'Deceptive Content' and section 3.10 'Snap Political and Advocacy Advertising Policies'

No label step. AI manipulation that misleads is banned.

Who must disclose

Any advertiser on Snapchat, including political and advocacy advertisers.

What triggers it

Ads that use generative AI (or deceptive editing) to manipulate content for false or misleading purposes. In political ads, also content with synthetic avatars, or visual or audio likenesses of a real person, manipulated for false or misleading purposes.

How

The policy describes a prohibition, not a label: no advertiser disclosure toggle or required wording for AI use is given in the policy text.

If you don't

Snap says all ads are subject to its review, that it may reject or remove any ad in its sole discretion, and that it may suspend or terminate accounts that violate the Advertising Policies.

For an agency

On Snapchat, AI-made creative is allowed as long as it is not false or misleading, and fake or manipulated likenesses of real people are off limits; there is no AI label step to complete.

Sources: Snap Advertising Policies

Pinterest Help, 'Gen AI labels' (section 'AI labels on ads')

No duty. Pinterest labels ads it detects as made with AI.

Who must disclose

Pinterest applies the label. Its wording depends partly on what the content owner (the advertiser) has indicated about AI use.

What triggers it

Ad Pins that Pinterest's systems detect were generated or modified with AI, or that the content owner indicated were modified with AI or contain an AI-generated person.

How

For ads the label does not show on the Pin close-up. It shows in the ellipsis (three-dot) menu next to the Pin, with the other details about the ad. The disclosure says the Pin was modified with AI, or contains an AI-generated person, based on what the content owner indicated, or that it may be modified with AI if Pinterest's systems flagged it.

If you don't

Pinterest adds the label itself. The page names no penalty and offers an appeal if you find an AI label on one of your Pins.

For an agency

Pinterest may label client ads as AI modified without being asked, so mark AI use honestly where Pinterest lets you and appeal a wrong label through Pinterest support.

Sources: Gen AI labels; Introducing Gen AI Labels: Pinterest is Taking a New Step in Transparency (2025-04-30)

Federal Trade Commission Act, Section 5(a), 15 U.S.C. 45 (no FTC rule requires labeling AI-made content) · United States

No federal rule requires an AI label. An AI-made person, result or claim that misleads is still deceptive.

Who must disclose

Any advertiser or marketer in or affecting commerce, including an agency that makes the ad or claim.

What triggers it

Unfair or deceptive acts or practices. The FTC has not adopted a rule that requires AI-made content to be labeled as AI. It applies the general deception rule to AI: for example it alleged that an AI-text-detector seller's 98 percent accuracy claim was false or unsubstantiated (Workado, proposed order published May 2025), and in July 2026 it proposed a policy statement saying that steering an AI system's outputs without telling consumers may deceive them.

How

No AI label mechanism exists in federal law. Whether a disclosure is needed depends on whether the content, without it, would mislead consumers about something material (for example a person, result or testimonial that was invented).

Not required

Nothing in Section 5 requires disclosing that a post, image or ad was made with AI where that fact does not make the message misleading.

If you don't

The FTC can seek orders and injunctions against deceptive practices; money remedies for a plain Section 5(a) violation are limited, and civil penalties are not available for it. Civil penalties need a rule violation (such as 16 CFR Part 465) or a violation of an order.

For an agency

No federal rule makes you label AI-made content as AI, but an AI-made person, result, review or claim that misleads consumers is a deception risk with or without an AI-specific rule.

Sources: 15 U.S.C. 45, Unfair methods of competition unlawful; Workado, LLC, f/k/a Content at Scale AI; Analysis of Proposed Consent Order To Aid Public Comment (2025-05-01); Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems (proposed) (2026-07-07); Trade Regulation Rule on the Use of Consumer Reviews and Testimonials (final rule, full text) (2024-08-22)

Trade Regulation Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465 (final rule published 22 August 2024) · United States

No label fixes it: AI-generated reviews and invented customers are banned outright.

Who must disclose

Any business that writes, creates, sells, buys or spreads reviews, consumer testimonials or celebrity testimonials about a business or its products. A marketing agency that writes or posts testimonials for a client is a business under the rule.

What triggers it

A review or testimonial that materially misrepresents that the reviewer or testimonialist exists, that they used the product or service, or what their experience was. The FTC says AI-generated reviews are covered. The rule has no AI label requirement: a fake reviewer (including an AI-invented person presented as a real customer) is banned outright.

How

There is no AI label to add. The duty is not to create, buy or spread such reviews and testimonials. The rule's own clear-and-conspicuous disclosure duty concerns insider reviews and testimonials (by officers, managers, employees or agents) under 16 CFR 465.5, not AI use.

Not required

A business is not liable under the buying, procuring and spreading provisions (465.2(b) and (c)) for reviews that came from generalized solicitations to purchasers to post their experiences, or for merely hosting reviews on its site.

If you don't

The FTC can seek civil penalties for rule violations (maximum $53,088 per violation under 16 CFR 1.98 as published after the 17 January 2025 adjustment; the figure is adjusted each year) and consumer redress. The FTC sent warning letters to 10 companies on 22 December 2025 under this rule.

For an agency

Do not publish AI-generated customers, reviews or testimonials as if they were real people who used the client's product; the rule bans that regardless of any AI label.

Sources: 16 CFR Part 465, Rule on the Use of Consumer Reviews and Testimonials (eCFR); Trade Regulation Rule on the Use of Consumer Reviews and Testimonials (final rule, 89 FR, full text) (2024-08-22); 16 CFR 1.98, Adjustment of civil monetary penalty amounts; FTC Issues Warning Letters Over Consumer Review Rule: What Marketers Need To Know (2026-01)

N.Y. General Business Law § 396-b, as amended by S.8420-A / A.8887-B (Laws of New York 2025, chapter 617, signed 11 December 2025) · New York

Yes. An ad showing an AI-made person who is not a real performer must say so conspicuously.

Who must disclose

Any person engaged in the business of dealing in property or services who, for a commercial purpose, produces or creates an advertisement, where that person has actual knowledge that a synthetic performer is in it.

What triggers it

An advertisement that contains a 'synthetic performer': a digitally created asset made or changed by computer using generative AI or a software algorithm, meant to give the impression of a human performer in an audiovisual or visual performance, who is not recognizable as any identifiable natural performer.

How

A conspicuous disclosure inside the advertisement that a synthetic performer is in it. The statute does not define 'conspicuously' and does not set wording, size, duration or placement.

Not required

Audio-only advertisements; AI used solely to translate the language of a human performer; advertisements and promotional materials for expressive works (films, TV, streaming, documentaries, video games) when the synthetic performer is used the same way in the advertisement as in the work. The medium that publishes the ad (newspaper, TV network, streaming service, billboard) is not covered by the section.

If you don't

Civil penalty of $1,000 for a first violation and $5,000 for each later violation. The statute does not create a private right of action and does not name who collects the penalty.

For an agency

If an ad you make for a client shows a computer-generated person who is not a real, recognizable performer, put a conspicuous disclosure in the ad itself (audio-only ads are exempt); AI used only for backgrounds, retouching of a real performer or text generally does not create a synthetic performer.

Sources: Assembly Bill A8887B (same as S8420-A): summary, actions (signed, chapter 617, 12/11/2025) and text (2025-12-11); N.Y. General Business Law § 396-B, Advertisements (consolidated text); Assembly Bill A11683 (2025-2026): amends GBL 396-b, status In Assembly Committee (2026-09-02); Client Alert: New York's Synthetic Performer Disclosure Law is Now in Effect (secondary source that the law took effect 9 June 2026) (2026-08-31)

California Government Code § 84514 (AB 2355, Statutes of 2024, chapter 260) · California

Yes. Committee ads with AI-made or materially AI-altered media must carry a set sentence.

Who must disclose

A committee (a political committee as defined in Government Code § 82013) that creates, originally publishes or originally distributes a qualified political advertisement. An agency may make the ad for the committee, but the statute puts the duty on the committee.

What triggers it

A political advertisement that contains an image, audio or video that is generated or substantially altered using AI: either entirely created by AI and would falsely appear authentic to a reasonable person, or materially altered by AI so that a reasonable person would understand it fundamentally differently from the unaltered version.

How

The ad must include, in a clear and conspicuous manner, the words 'Ad generated or substantially altered using artificial intelligence.', displayed or spoken in the manner California's political-ad disclosure sections (Gov. Code §§ 84504 to 84504.5) prescribe for the medium.

Not required

Media that is only immaterially altered by AI, including cosmetic adjustments, color edits, cropping or resizing.

If you don't

The Fair Political Practices Commission may seek an injunction or pursue administrative or civil remedies; a violation is not a misdemeanor.

For an agency

If you produce a California committee's ad with AI-generated or materially AI-altered image, audio or video, make sure the committee's ad carries the exact AI disclosure sentence in the format the medium requires.

Sources: California Government Code § 84514, Disclosure in Advertisements

State statutes regulating deepfakes or AI-altered media in election communications, as counted by Public Citizen's state tracker (30 states on 13 May 2026), Ballotpedia (31 states on 3 April 2026) and NCSL · United States

Usually yes: about 30 states regulate AI deepfakes in election ads, by label or by ban.

Who must disclose

Varies by state: usually whoever distributes or pays for the political ad or election communication (candidates, committees, sometimes any person); some states also reach platforms (blocked in California).

What triggers it

Realistic AI-generated or AI-altered audio, images or video of a candidate or election matter in political ads or election communications, usually in a window before an election. States split between disclosure laws (a required label on AI-altered political ads) and outright bans on deceptive deepfakes, with exceptions for satire and parody; details differ in each state.

How

Set by each statute: typically a stated disclaimer on the ad (for example California requires 'Ad generated or substantially altered using artificial intelligence.' on committee ads, and 'This [image/audio/video] has been manipulated.' under its election-deepfake section). Check each state where the ad runs.

Not required

Commonly satire and parody, and minor edits such as color or cropping, but this varies; check the state text.

If you don't

Varies: injunctions, civil penalties, private lawsuits and in some states criminal penalties. Courts have blocked laws in some states (Public Citizen's tracker marks Hawaii's 2024 law 'Struck down'; California's AB 2839 was struck down in 2025 and then rewritten in 2026).

For an agency

For any political ad with AI-made or AI-altered people, voices or scenes, check the statute of every state where it will run, because wording, timing windows and penalties differ; no federal rule applies.

Sources: Tracker: State Legislation on Deepfakes in Elections (last updated 21 October 2026) (2026-10-21); 30 States Now Have Laws to Regulate Election Deepfakes (2026-05-13); 15 deepfake bills enacted so far this year, number of states with deepfake laws remains the same (2026-04-03)

California Business and Professions Code §§ 22757-22757.6 (SB 942 of 2024; amended by AB 853 of 2025 and SB 1000 of 2026, chapter 861, effective 30 September 2026) · California

Not for agencies. AI tool makers must embed hidden provenance data in images, video and audio.

Who must disclose

'Covered providers': a person that creates, codes or otherwise produces a generative AI system that is publicly accessible in California. SB 1000 deleted the earlier 1,000,000-monthly-user threshold. Later duties fall on large online platforms and GenAI hosting platforms (from 1 January 2027) and on capture-device makers (from 1 January 2028). It does not bind a company that only uses an AI tool.

What triggers it

Image, video and audio content (not text) created or altered, other than by minor modification, by the provider's generative AI system.

How

The provider must embed a latent (hidden, machine-readable) disclosure that names the provider, the system and version, the time and date, a unique identifier and whether the system created or altered the content, and that is permanent or extraordinarily difficult to remove; it must also offer a free public tool that checks content against that disclosure. SB 1000 deleted the earlier duty to offer users an option to add a visible label.

Not required

Minor modifications (brightness, contrast, color, sharpening, resizing, scaling, cropping, format conversion, audio denoising); products that are exclusively non-user-generated video games; before 1 January 2029, GenAI systems designed primarily as assistive technology.

If you don't

$5,000 per violation, each day a separate violation, in a civil action by the Attorney General, a city attorney or a county counsel, plus attorney's fees; $50,000 per violation for falsely claiming a system is assistive technology.

For an agency

The Act puts the duty on the companies that build generative AI tools, not on an agency using them, so it does not make you add a label; expect AI images, video and audio from major tools to carry hidden provenance data, and from 2027 large platforms must show it.

Sources: California Business and Professions Code, Division 8, Chapter 25, AI Transparency Act (§§ 22757-22757.6); SB 1000 (Becker), chapter 861, Statutes of 2026, California AI Transparency Act (chaptered 30 September 2026) (2026-09-30)

California Business and Professions Code §§ 17940-17943 (SB 1001 of 2018) · California

Yes, for bots that sell or sway votes in California: say it is a bot.

Who must disclose

Any person who uses a bot to communicate or interact online with someone in California.

What triggers it

Using a bot with the intent to mislead the other person about its artificial identity, to knowingly deceive them about the content of the communication in order to encourage a purchase or sale or to influence a vote. A 'bot' is an automated online account where all or substantially all actions or posts are not the result of a person.

How

Tell the person it is a bot. The disclosure must be clear, conspicuous and reasonably designed to inform people who communicate or interact with the bot that it is a bot. A person who discloses is not liable under the section.

Not required

Online platforms and service providers such as web hosts carry no duty under it. A bot is an account where all or substantially all of the posts are not the result of a person, so an account a person runs, even with AI-drafted posts, is not clearly a bot.

If you don't

The chapter names no penalty or enforcer of its own and says its duties add to those of other laws.

For an agency

If you run an automated account or chatbot for a client that talks to Californians to sell something or sway a vote, tell them it is a bot; an account a person actually operates, even with AI-drafted posts, is not clearly a 'bot' under this definition.

Sources: California Business and Professions Code, Division 7, Part 3, Chapter 6, Bots (§§ 17940-17943)

Utah Code Title 13, Chapter 77, Generative Artificial Intelligence: Consumer Disclosures and Enforcement (enacted by SB 226, Laws of Utah 2025, chapter 465; definitions amended by chapter 95, 2026 General Session), building on the Artificial Intelligence Policy Act of 2024 (Utah Code Title 13, Chapter 72) · Utah

Yes for chatbots when a person asks, and up front in licensed high-risk services.

Who must disclose

A 'supplier' that uses generative AI to interact with an individual in a consumer transaction; and a person providing services in an occupation that the Utah Department of Commerce regulates and that requires a license or state certification, when using generative AI.

What triggers it

Generative AI that simulates human conversation by text, audio or visual communication and produces non-scripted, human-like output with limited or no human oversight (a chatbot or voice agent, not AI-written copy or images). A supplier must disclose it is AI when the person clearly and unambiguously asks whether they are dealing with a human or AI. A licensed professional must disclose up front when using generative AI in a 'high-risk' interaction: collecting health, financial or biometric data, or giving personalized financial, legal, medical or mental health advice that a person could rely on for significant decisions.

How

Supplier: say it is generative AI and not a human when asked. Licensed professional: disclose verbally at the start of a spoken interaction and in writing before a written one. Safe harbor: no enforcement action if the AI clearly and conspicuously says, at the start and throughout, that it is generative AI, not human, or an AI assistant.

Not required

Outside licensed occupations there is no duty to volunteer AI status unless the person clearly and unambiguously asks. Content that is not a simulated human conversation (AI-written ads, images, video) is outside the definition.

If you don't

A violation counts as a violation of the Utah Consumer Sales Practices Act (13-11-4(1)). The Division of Consumer Protection may impose an administrative fine of up to $2,500 per violation and sue in court (court fine up to $2,500 per violation, injunction, disgorgement, fees); violating an order carries up to $5,000 per violation.

For an agency

If you build or run a chatbot or voice agent for a client that talks to Utah consumers, it must say it is AI when asked (or label itself up front for the safe harbor); this law does not cover AI-written copy, images or video.

Sources: Utah Code 13-77-103, Required disclosures; Utah Code 13-77-101, Definitions; Utah Code 13-77-104, Safe harbor; Utah Code 13-77-105, Enforcement; Utah Code Title 13, Chapter 72, Artificial Intelligence Policy Act (index, 'Affected by 63I-2-213 on 7/1/2027'); New Utah AI Laws Change Disclosure Requirements and Identity Protections, Target Mental Health Chatbots (secondary summary of the 2025 amendments)

10 M.R.S. § 1500-DD, Required disclosure of use of artificial intelligence chatbot to engage in trade and commerce (P.L. 2025, chapter 294, LD 1727; reallocated from § 1500-Y) · Maine

Yes, for a chatbot a reasonable consumer could mistake for a person.

Who must disclose

A person who uses an AI chatbot or any other computer technology to engage in trade and commerce with a consumer.

What triggers it

Using a chatbot or other computer technology with a consumer in a way that may mislead or deceive a reasonable consumer into believing they are dealing with a human being. An 'artificial intelligence chatbot' is a software application, web interface or program that simulates human conversation through text or voice.

How

Notify the consumer, in a clear and conspicuous manner, that they are not engaging with a human being.

Not required

The text prohibits only use that may mislead a reasonable consumer into thinking the consumer is dealing with a human; it lists no other exemptions.

If you don't

A violation is a violation of the Maine Unfair Trade Practices Act.

For an agency

A chatbot or voice agent you run for a client that talks to Maine consumers must tell them it is not a human unless no reasonable consumer could think it is one.

Sources: 10 M.R.S. § 1500-DD (Maine Revised Statutes); LD 1727 / HP 1154, An Act to Ensure Transparency in Consumer Transactions Involving Artificial Intelligence (final disposition, chapter 294); Maine Legislature home page (general effective dates for the 132nd Legislature's sessions); 2026 State Chatbot Laws: Key Provisions and Regulatory Trends (secondary overview of other states' chatbot laws) (2026-04-29)

Colorado SB 26-189, Concerning the Use of Automated Decision-Making Technology in Consequential Decisions (C.R.S. 6-1-1701 and following), signed 14 May 2026; repeals and reenacts SB 24-205 (Colorado AI Act) · Colorado

Not for marketing content. From January 1, 2027, a notice applies to automated decisions about jobs, housing, credit and similar.

Who must disclose

Deployers (and developers) of automated decision-making technology that materially influences a 'consequential decision' about a person: education, employment, housing, financial or lending services, insurance, health-care services, or essential government services.

What triggers it

Use of such technology in those decisions. It is not a rule about AI-made marketing content or general chatbots.

How

A clear and conspicuous notice to the consumer at the point of interaction that the technology is used in a consequential decision (a prominent public notice near the point of interaction satisfies it), and within 30 days after an adverse outcome a plain-language description of the technology's role and how to get more information. The Attorney General is to adopt clarifying rules by 1 January 2027.

Not required

Decisions outside the listed consequential areas; entities that comply with other laws to the extent the act exempts them.

If you don't

A violation is a deceptive trade practice enforced by the Attorney General under the Colorado Consumer Protection Act, with a 60-day notice and chance to cure before any action begun before 1 January 2030. No new private right of action.

For an agency

This does not require labeling marketing content or chatbots; it matters only if a client uses automated tools to make decisions about people's jobs, housing, credit, insurance, health care or education, and the 2024 act's older 'tell consumers they are interacting with AI' duty is gone from the replacement.

Sources: SB26-189, Automated Decision-Making Technology (bill page with summary as enacted; status Became Law, signed act dated 05/14/2026) (2026-05-14); Senate Bill 26-189, signed act (PDF; section 5: effective 1 January 2027, applies to consequential decisions made on or after that date) (2026-05-14); Colorado AI law in flux: Comprehensive replacement bill signed after federal court blocks predecessor's enforcement (secondary source)

FCC Notice of Proposed Rulemaking, Implications of Artificial Intelligence Technologies on Protecting Consumers From Unwanted Robocalls and Robotexts (CG Docket No. 23-362, FCC 24-84, published 10 September 2024), plus existing rule 47 CFR 64.1200 · United States

Proposed only. AI voice calls already need the consent that artificial-voice calls need.

Who must disclose

Anyone placing outbound marketing calls or texts, including an agency running a phone or SMS campaign for a client.

What triggers it

Proposed: a caller using an AI-generated voice would have to tell the person, at the start of each call, that the call uses AI-generated technology, and consent language would have to say that consent covers AI-generated calls and texts. Already in force: the FCC treats AI voices as an 'artificial or prerecorded voice' under the TCPA, so the consent and caller-identification rules for artificial voices apply.

How

Proposed: a spoken statement at the beginning of each AI-voice call and a disclosure in the consent request. In force today: the message must begin by stating clearly the identity of the business responsible for the call.

Not required

The proposal would also exempt AI uses that help people with disabilities use the phone network.

If you don't

Calls that break the TCPA's artificial-voice rules carry TCPA liability. No penalty attaches to the AI-disclosure rule because it was not adopted.

For an agency

There is no FCC rule yet that makes you say a call is AI-generated, but AI voice calls already need the consent that artificial-voice calls need, and the caller's identity must be stated at the start.

Sources: Implications of Artificial Intelligence Technologies on Protecting Consumers From Unwanted Robocalls and Robotexts (proposed rule) (2024-09-10); 47 CFR 64.1200, Delivery restrictions (eCFR)

Regulation (EU) 2024/1689 (AI Act), Article 50(4) first subparagraph and Article 50(5) · European Union

Yes. Whoever uses AI to make a realistic deepfake for professional use must label it visibly or audibly.

Who must disclose

The deployer of the AI system: any person or organization that uses an AI system under its own authority for professional (not purely personal) activity. The Commission's guidelines say an advertising company that uses the system is the deployer, even when freelancers or contractors operate it for the company, and that a client who only commissions an ad, without deciding whether or how the agency uses AI, is not a deployer. A deployer outside the EU is covered when it foresees the content being used in the EU.

What triggers it

Image, audio or video content made or altered with AI that is a deep fake: it resembles existing people, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. The Commission's guidelines list as deep fakes a realistic AI depiction of a celebrity influencer in an ad and an AI product image that makes the product look different from, or better than, the real one. They list as not deep fakes a flying sphinx, cartoon mice in an ad, and a real product shown against an AI-generated background when the ad does not mislead about the product.

How

A clear and distinguishable disclosure that the content was artificially generated or manipulated, given to the person at the latest on first exposure, perceivable without special tools (for example a visible or audible label) and meeting accessibility requirements. The Commission says the provider's machine-readable marking does not count as the deployer's disclosure. The voluntary Code of Practice on Transparency of AI-generated Content (assessed as adequate by the Commission and the AI Board) can be signed to show compliance, and the Commission publishes EU icons that may be used as the label.

Not required

Content that is not a deep fake under the three tests (resemblance, real or plausible subject, false appearance of authenticity). Light edits such as colour correction, background tidying, re-scaling or extending a background, per the guidelines. Personal non-professional use. Use authorised by law for crime detection and prosecution. Content made before 2 August 2026 does not have to be labelled retroactively. Where the work is evidently artistic, creative, satirical or fictional, the duty shrinks to a disclosure that does not hamper display or enjoyment of the work.

If you don't

Administrative fines up to EUR 15 million or 3% of total worldwide annual turnover for the preceding year, whichever is higher (for SMEs and small mid-cap companies, whichever is lower). Enforced mainly by national market surveillance authorities; each Member State sets its own penalty rules.

For an agency

If you use AI to make realistic images, video or voice of real or believable people, places, products or events for a client's audience in the EU, you are the deployer and must label them visibly or audibly at first exposure, unless the content is clearly fantasy, only lightly edited, or an evidently creative work that needs only a light-touch label.

Sources: Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 3(60), 50, 99 and 113 (2024-07-12); Regulation (EU) 2026/1744 amending Regulation (EU) 2024/1689 (Digital Omnibus on AI) (2026-07-24); Transparency obligations under Article 50 of the AI Act (questions and answers) (2026-07-24); Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of Regulation (EU) 2024/1689, C(2026) 5054 final (annex) (2026-07-20); Guidelines on transparency obligations for providers and deployers of certain AI systems (2026-08-06); Code of Practice on Transparency of AI-generated Content (2026-10-19)

Regulation (EU) 2024/1689 (AI Act), Article 50(4) second subparagraph and Article 50(5) · European Union

Yes, unless a person reviewed it and someone holds editorial responsibility. Ordinary ad copy is outside it.

Who must disclose

The deployer of the AI system that publishes the text (the same meaning of deployer as for deep fakes: a person or organization using an AI system under its authority for professional activity).

What triggers it

Text that is generated or manipulated by AI and published with the purpose of informing the public on matters of public interest. The Commission says all three must be true: the text is published, it informs the public, and it concerns public-interest matters such as politics, public services, justice, fundamental rights, public security, public health, environmental protection, consumer safety, or economic, financial, scientific or cultural developments that may be a subject of public debate.

How

A clear and distinguishable label that the text was artificially generated or manipulated, given at the latest at first exposure and meeting accessibility requirements. The deployer part of the voluntary Code of Practice on Transparency of AI-generated Content covers this case.

Not required

Text that has gone through a process of human review or editorial control where a person or organization holds editorial responsibility for the publication. The Commission says spell-checking or grammar correction is not enough: human review means a person with relevant knowledge deliberately examining the substance, and editorial responsibility means someone holds the ultimate legal responsibility for the publication. The guidelines list AI-written text that is part of a company's advertisement or product descriptions (without claims on health, consumer safety or sustainability) as outside this rule, along with AI-generated fiction and text only shown to the person who prompted it.

If you don't

Administrative fines up to EUR 15 million or 3% of total worldwide annual turnover for the preceding year, whichever is higher (for SMEs and small mid-cap companies, whichever is lower). Enforced mainly by national market surveillance authorities.

For an agency

AI-written articles or posts on public-interest topics (health, politics, public services, safety, finance and similar) published for an EU audience need an AI label unless a named person or editor has really reviewed and approved them, while the Commission treats ordinary ad copy and product descriptions as outside this rule.

Sources: Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 3(60), 50, 99 and 113 (2024-07-12); Transparency obligations under Article 50 of the AI Act (questions and answers) (2026-07-24); Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of Regulation (EU) 2024/1689, C(2026) 5054 final (annex) (2026-07-20); Guidelines on transparency obligations for providers and deployers of certain AI systems (2026-08-06); Code of Practice on Transparency of AI-generated Content (2026-10-19)

Regulation (EU) 2024/1689 (AI Act), Article 50(1) and Article 50(5) · European Union

Yes. People must be told they are talking to an AI from the first message, unless it is obvious.

Who must disclose

The provider of an AI system meant to interact directly with people (chatbots, AI agents, avatars): whoever develops it, or has it developed, and puts it on the market or into service under its own name or trademark.

What triggers it

AI systems designed for a genuine two-way exchange in which the AI itself communicates directly with a natural person, such as a customer service chatbot or AI avatar.

How

The system must be designed and built so that the person is told they are interacting with an AI system, clearly and distinguishably, from the start of the first interaction, and in line with accessibility requirements.

Not required

When it is obvious to a reasonably well-informed, observant and circumspect person, given the circumstances and context of use (the Commission says to read this exception narrowly). Systems that run only in the background, machine-to-machine, or without direct contact with people. Systems authorised by law for criminal law enforcement, unless open to the public to report a crime.

If you don't

Administrative fines up to EUR 15 million or 3% of total worldwide annual turnover for the preceding year, whichever is higher (for SMEs and small mid-cap companies, whichever is lower). Enforced mainly by national market surveillance authorities.

For an agency

If you build or launch a branded chatbot or AI avatar for a client's EU audience, the chat must tell people it is an AI from the first message unless that is obvious, and you and the client should settle in the contract which of you is the provider.

Sources: Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 3(60), 50, 99 and 113 (2024-07-12); Transparency obligations under Article 50 of the AI Act (questions and answers) (2026-07-24); Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of Regulation (EU) 2024/1689, C(2026) 5054 final (annex) (2026-07-20); Guidelines on transparency obligations for providers and deployers of certain AI systems (2026-08-06)

Regulation (EU) 2024/1689 (AI Act), Article 50(2), with the transitional rule in Article 111(4) added by Regulation (EU) 2026/1744 · European Union

Not for agencies that use AI tools. It binds the companies that make them (to December 2, 2026 for existing tools).

Who must disclose

Providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video or text: the companies that build and offer the generator, not the agencies or clients that use it. A provider is whoever develops an AI system, or has it developed, and puts it on the market or into service under its own name or trademark.

What triggers it

The outputs of the generative AI system must be marked in a machine-readable format and be detectable as artificially generated or manipulated.

How

Technical marks built into the tool, such as metadata or watermarks, that are effective, interoperable, robust and reliable as far as technically feasible. Section 1 of the voluntary Code of Practice on Transparency of AI-generated Content sets out measures for this. For systems already on the market before 2 August 2026, providers have until 2 December 2026.

Not required

AI systems that only perform an assistive function for standard editing or do not substantially alter the input data or its meaning. The Commission's Q&A also puts outside the duty short strings of characters, source code, machine-to-machine outputs never shown to people, and outputs used only in closed-loop industrial or product-development settings unless they are the final output. A narrow exemption exists for some business-to-business or industrial uses.

If you don't

Administrative fines up to EUR 15 million or 3% of total worldwide annual turnover for the preceding year, whichever is higher (for SMEs and small mid-cap companies, whichever is lower). Enforced mainly by national market surveillance authorities; for some systems built on general-purpose models, by the AI Office.

For an agency

An agency that only uses a vendor's AI tool does not carry this duty (it carries the visible-label duty in Article 50(4) instead), and the tool's embedded marks do not replace your own label, but an agency that offers its own AI generator under its own name becomes a provider.

Sources: Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 3(60), 50, 99 and 113 (2024-07-12); Regulation (EU) 2026/1744 amending Regulation (EU) 2024/1689 (Digital Omnibus on AI) (2026-07-24); Transparency obligations under Article 50 of the AI Act (questions and answers) (2026-07-24); Code of Practice on Transparency of AI-generated Content (2026-10-19); Commission publishes Code of Practice on marking and labelling AI-generated content (2026-06-10); Strong backing for the Code of Practice on Transparency of AI-generated Content (2026-07-31)

No AI-specific statute. The UK Code of Non-broadcast Advertising (CAP Code) and the BCAP Code apply to AI-made ads like any other ad (rules enforced by the Advertising Standards Authority) · United Kingdom

No law requires an AI label. Ads must still not mislead, and a label does not fix a misleading ad.

Who must disclose

No one has a legal duty to label AI content as such. Advertisers (and the agencies and publishers who make and place the ads) must meet the existing CAP Code rules, for example on misleading ads, whatever tool made the content.

What triggers it

Ads that mislead, or are harmful or irresponsible, are banned however they were made. The ASA's guidance says an ad using AI breaks the rules if the AI-made content misleads (for example an AI image of a cosmetic's effect that does not reflect real results, or a deepfake celebrity endorsement that looks real).

How

No AI label is required. The CAP guidance suggests asking two questions: would the audience be misled if AI use is not disclosed, and would a disclosure clarify the ad's message or contradict it. It points to the 2023 ISBA and IPA principle that AI use should be transparent where it features prominently in an ad and is unlikely to be obvious.

Not required

The ASA notes that advertisers have long used post-production and editing without having to disclose it, and says AI changes that make only subtle non-material edits, or that replace actors, photographers or CGI, change little in how consumers see an ad.

If you don't

Complaints about ads go to the Advertising Standards Authority, which rules on them under the CAP Code. A disclosure that AI was used does not cure a misleading ad: the ASA says disclosure alone is very unlikely to fix a fundamentally misleading message.

For an agency

UK law does not make you label AI-made ads, but the ads must still be truthful and responsible, and you cannot fix a misleading AI image by adding an 'AI' label.

Sources: AI content labelling (Research Briefing CBP-10467) (2026-01-20); Disclosure of AI in Advertising: Striking the Balance Between Creativity and Responsibility (2025-05-29); Regulating the Illusion of Intelligence in Ads (2026-08-27); AI and Deepfakes: Four Things Advertisers Need to Know Before They Hit "Run" (2026-06-11)

人工智能生成合成内容标识办法 (Measures for Labeling AI-Generated Synthetic Content), notice 国信办通字〔2025〕2号 of the Cyberspace Administration of China, the Ministry of Industry and Information Technology, the Ministry of Public Security and the National Radio and Television Administration; with the mandatory national standard GB 45438-2025 · China

Yes. Users must declare AI-made content when they publish it and use the platform's label.

Who must disclose

Online information service providers that offer AI generation or synthesis services (they add the labels), platforms that spread content (they check and flag it), and app stores (they ask apps whether they offer AI generation and check their labeling materials). Users who publish AI-generated content on a platform must declare it and use the platform's labeling function (Article 10). The Measures apply to services covered by China's algorithm recommendation, deep synthesis and generative AI rules.

What triggers it

Text, images, audio, video and virtual scenes generated or synthesized with AI (Article 3).

How

Explicit labels that people can see or hear (words, symbols, sound or graphics): for text, a notice at the start, end or middle or around the text; for audio, a spoken or rhythm cue or an on-screen notice; for images, a prominent notice in a suitable place; for video, a prominent notice on the first frame and around playback. Implicit labels written into file metadata (content attributes, provider name or code, content ID). Platforms must check metadata and add a prominent notice to content that has an implicit label, that the user declares as AI-made, or that they suspect is AI-made. The mandatory national standard GB 45438-2025 sets the detailed methods.

Not required

A service may hand over content without an explicit label if the user asks for it and accepts the labeling duty and responsibility in the user agreement, and the service keeps logs for at least six months (Article 9); the CAC says the user must still declare it and add an explicit label before publishing to the public. Where a specific field such as news, film or creative work has its own labeling rules, those prevail (CAC Q&A).

If you don't

Handled by the internet information, telecom, public security and radio and TV authorities under existing laws, regulations and rules (Article 13); the Measures state no fine amount. No one may maliciously delete, alter, forge or hide the labels, or provide tools for doing so (Article 10).

For an agency

For content published on Chinese platforms, declare AI-made content when you post and use the platform's AI label function, and do not strip labels or metadata the AI tool added.

Sources: 关于印发《人工智能生成合成内容标识办法》的通知 (2025-03-07); 《人工智能生成合成内容标识办法》答记者问 (Q&A on the Measures) (2025-03-14); GB 45438-2025 网络安全技术 人工智能生成合成内容标识方法 (Cybersecurity technology: method for labeling AI-generated synthetic content) (2025-02-28)

Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (AI Basic Act), Act No. 20676, Article 31, with its Enforcement Decree and the Ministry of Science and ICT Guidelines on Ensuring AI Transparency · South Korea

Not for agencies that only use AI tools. It binds AI service operators, with penalties deferred for at least a year.

Who must disclose

An 'AI business operator' that provides generative AI or a product or service using it to users in Korea, including foreign companies that serve Korean users. The ministry's guidelines say people and companies that only use AI tools for work or creative activities are users, not operators, and are not subject to the duty (its example: a film company that makes movies with AI video tools).

What triggers it

Output from generative AI. Operators must tell users in advance that a product or service runs on generative AI (Article 31(1)), label that output was generated by generative AI (Article 31(2)), and clearly notify or label virtual sound, image or video that is difficult to distinguish from the real thing (Article 31(3)).

How

Per the ministry's guidelines: inside the service, flexible methods such as notices or visual indicators in the interface; when output is downloaded or shared outside the service, a visible or audible identifier such as a watermark, or machine-readable metadata after a text or voice notification; and for deepfake-type output, a label that users can clearly recognize. Artistic or creative works may be labeled in a way that does not hinder display or enjoyment (Article 31(3)).

Not required

Users who simply use AI as a work or creative tool (ministry guidelines). Outputs that stay inside the service environment can use lighter labeling than outputs that are exported. For artistic or creative works the label may be placed so it does not hinder exhibition or enjoyment (Article 31(3)).

If you don't

The Minister of Science and ICT can investigate suspected breaches of Article 31(2) and (3) and order the operator to stop or correct them (Article 40). Administrative fines of up to 30 million won apply to failing to give the advance notice under Article 31(1), to failing to designate a domestic agent, and to ignoring a corrective order (Article 43); the fine list does not name Article 31(2) or (3) directly. The ministry says investigations and penalties under the transparency provision are deferred for a grace period of at least one year from 22 January 2026.

For an agency

The Korean labeling duty sits with the AI tool or service provider, not with an agency that only uses AI to make content for clients, but an agency that offers its own AI generator or chatbot to Korean users would be an operator.

Sources: Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (Act No. 20676, 21 January 2025), English translation (2025-01-21); MSIT Releases Guidelines on Ensuring AI Transparency (2026-01-22); The AI Basic Act Comes into Force to Lay the Foundation for Korea to Become an AI G3 (2026-01-22)

Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 (G.S.R. 120(E), 10 February 2026), amending the IT Rules, 2021: new Rule 2(1)(wa), Rule 3(3) and Rule 4(1A) · India

Platforms must label realistic AI content and ask uploaders to declare it, so answer the declaration truthfully.

Who must disclose

Intermediaries: platforms and services that let users create, share or publish such content must label it and embed metadata (Rule 3(3)). Significant social media intermediaries must, before content is published, ask the uploader to declare whether it is synthetically generated, check the declaration with technical measures, and label confirmed content (Rule 4(1A)). So the direct legal duty is on the platform; the uploader (including an agency) meets it by answering the declaration truthfully.

What triggers it

'Synthetically generated information': audio, visual or audio-visual content that is artificially or algorithmically created or altered using a computer resource so that it appears real, authentic or true and depicts an individual or event in a way that is, or is likely to be taken as, indistinguishable from a real person or event (deepfakes, realistic AI images, video and voice).

How

Platforms add a prominent label that is easily noticeable in the visual display (or a prefixed audio disclosure for audio) and embed permanent metadata or another provenance mechanism with a unique identifier where technically feasible, and they must not let anyone remove the label or metadata. Large social platforms add a declaration step at upload, verify it and show a label. MeitY's FAQ gives the example of an 'AI-generated: Yes/No' prompt.

Not required

Routine or good-faith editing, formatting, enhancement, technical correction, colour adjustment, noise reduction, transcription or compression that does not materially change the substance or meaning; good-faith creation of documents, presentations, training material and illustrative or hypothetical content that does not make a false document or record; and uses solely to improve accessibility, clarity, translation or searchability (Rule 2(1)(wa) proviso). Text-only AI output is not synthetically generated information under MeitY's FAQ. Content that breaks the law (for example child abuse material, non-consensual intimate images, false documents, deceptive depiction of a real person or event) must be blocked, not merely labeled.

If you don't

An intermediary that does not follow the IT Rules loses its exemption from liability for third-party content under section 79 of the IT Act and can face action under other laws (PIB, 6 August 2026). For a significant social media intermediary that knowingly permits or fails to act on unlabeled synthetic content, Rule 4(1A) deems it to have failed due diligence.

For an agency

When you upload realistic AI-made video, images or audio to a large platform for India, expect a declaration prompt and a label: answer truthfully, and do not strip labels or metadata from files.

Sources: Frequently Asked Questions on the IT (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026 (2026-02-10); The Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 [updated as on 10.02.2026] (2026-02-10); Government Strengthens Regulatory Framework to Address AI-Generated Deepfakes (2026-08-06); Notice: extension of time for feedback on draft amendments to the IT Rules, 2021 (including Rule 3(3)(a)(ii)) (2026-04-21); Exclusive: Amid protests, MeitY widens labelling for social media beyond synthetic content (2026-10-08)