Your country-by-country compliance map for AI-generated ad creative in 2026.

If you’re running AI-generated ads on Meta, TikTok, or YouTube, your compliance exposure doesn’t stop at the US or EU border. At least nine more markets now have enforceable or near-enforceable AI advertising laws, and the approaches vary dramatically by region. 

China requires both visible labels and embedded metadata. India makes platforms liable for unlabeled synthetic content, and its ad industry body adds its own labeling rules in December. Turkey has required AI disclosure in ads since August, and South Korea requires a “virtual person” label on AI-generated endorsers. The UK, Canada and Australia apply existing consumer law to AI ads, with fines that keep growing.

We’ve already covered the US and EU regulations in separate posts. This one picks up where those left off: nine more markets, country by country, covering what the law says, who’s liable, what the penalties look like, and whether human creator content sidesteps the rules.

TL;DR:

UK: the existing rules approach

The UK has no AI-specific advertising law, and the ASA has no general rule requiring you to disclose AI use. But two existing systems apply to AI ads in full.

The CAP Code (ASA). In June 2026, the ASA set out its position: the Code applies however an ad is made. A deepfake endorsement realistic enough to mislead is likely to breach it, and the advertiser stays responsible for AI output even when an automated platform generated or placed the ad.

Consumer law (CMA). Since April 6, 2025, the Digital Markets, Competition and Consumers Act lets the CMA fine businesses directly for consumer law breaches, including fake reviews and hidden advertising. Fines go up to 10% of global turnover, or £300,000 if that’s higher. The CMA is using those powers: on October 14, 2026, it fined Wayfair, Appliances Direct and Gold’s Gym a combined £6 million over misleading sales tactics.

Let’s say you run a Meta campaign in the UK where an AI-generated “customer” shares her results. Under the CAP Code that’s a misleading testimonial, and under the DMCC Act it’s a fake review, whether or not you label it as AI.

The human UGC angle: Real creators sharing real experiences meet both rule sets, as long as the paid relationship is disclosed with #ad. If you’re casting for the UK, here’s how to find UGC creators in the UK.

Canada: the “self-regulation” model

Canada doesn’t have federal AI advertising laws. The previously proposed Artificial Intelligence and Data Act (AIDA) died on the order paper in January 2025 when Parliament was prorogued. In June 2026, the government tabled two bills that touch AI, and neither is law yet. Bill C-34, the Safe Social Media Act (June 10), would require large social media services to label synthetic content. Bill C-36 (June 15) adds transparency rules for automated decisions.

What Canada does have is a well-developed industry self-regulation system, backed by a statutory fallback that carries real penalties.

What The Ad Standards Guidelines Require

In October 2025, Ad Standards Canada released updated Influencer Marketing Disclosure Guidelines that specifically address AI-generated content and AI-created influencers. The guidelines are straightforward: if ad content is generated or significantly altered by AI, that fact should be disclosed. The recommended hashtags are #MadeWithAI or #AIcreated.

AI-generated influencers get their own set of rules. They must follow the same material connection disclosure requirements as human influencers, with best practice being a clear label like #VirtualInfluencer or #AIinfluencer. The Canadian Code of Advertising Standards also imposes a practical limitation: AI-generated influencers shouldn’t provide testimonials about product qualities they can’t experience, such as taste. Posts created using AI filters or tools must remain truthful about the product’s actual benefits and attributes.

For paid partnership disclosure, #Ad used on its own (not combined with other words like #Advert or #Sponsored) remains the “gold standard” hashtag, with #Gifted and #InvitedGuest now recognized as acceptable alternatives for those specific scenarios.

The Competition Act Backstop

Ad Standards is industry self-regulation, not government enforcement. Compliance depends on member commitments and a complaint resolution process rather than statutory penalties. That distinction matters if you’re thinking about how much weight these guidelines carry.

However, the Competition Act provides a statutory backstop with real teeth. False or misleading advertising is illegal under the Act, and penalties scale quickly: individuals face fines up to C$750,000 on a first offence and C$1 million on subsequent violations. Corporations face the greater of C$10 million (C$15 million for repeat orders), three times the benefit, or 3% of annual worldwide revenue if the benefit can’t be determined.

The Competition Bureau’s 2026-27 Annual Plan names AI as one of its focus areas, alongside deceptive marketing practices. So while the AI-specific disclosure guidelines are voluntary, the underlying prohibition on misleading advertising is not.

The human UGC angle: Real creators filmed on camera need only disclose the paid partnership with #Ad. No #MadeWithAI or #AIcreated disclosure is required. And unlike AI-generated influencers, real creators face no restrictions on testifying about product qualities, because they actually experienced the product.

Australia: the “consumer law” approach

Australia has taken a different path from most countries on this list. Rather than writing new AI-specific legislation, the government reviewed its existing consumer protection framework and concluded it already covers AI. The Australian Treasury’s October 2025 final report on AI and the Australian Consumer Law (ACL) found that the existing principles-based protections are “generally well-suited” for AI-enabled goods and services. The core finding: the ACL’s prohibition on misleading or deceptive conduct applies regardless of the technology used to create the content, so AI-generated advertising is already in scope.

Australia’s National AI Plan (December 2025) confirmed the direction: no standalone AI law for now, with existing laws doing the work.

ACCC Enforcement: From Warnings to Fines

The practical question for brands isn’t whether the law covers AI advertising (it does), but whether regulators are actually enforcing it. And on that front, the answer shifted decisively in 2026.

The ACCC’s 2026–27 priorities target manipulative and false practices in digital markets, including dark patterns. AI isn’t named as a separate priority, but it doesn’t need to be: claims about how much AI a product uses, and AI-generated content that misleads, fall under the same misleading conduct rules.

In March 2026, the ACCC issued its first influencer-related penalties: two infringement notices totalling A$39,600 to PhotobookShop, which had hidden that reviews came from gifted products. The ACCC is also taking Microsoft to the Federal Court (filed October 2025) for allegedly misleading about 2.7 million Australians over Microsoft 365 subscription changes. The fines are small so far, but the direction is clear.

Penalty framework

The financial exposure doubled on March 28, 2026, when the Treasury Laws Amendment (Doubling Penalties for ACCC Enforcement) Act 2026 took effect. Maximum ACL penalties for corporations now reach the greater of A$100 million (up from A$50 million), three times the benefit attributable to the contravention, or 30% of adjusted turnover during the relevant period. That’s well above the EU AI Act’s cap for transparency breaches (€15 million or 3% of turnover).

Two other developments are worth flagging: the Commercial Radio Code of Practice now requires disclosure when a synthetic AI voice hosts a regular program or news (since July 1, 2026), and Privacy Act amendments taking effect December 10, 2026 will require disclosure when automated systems use personal data to make decisions that significantly affect individuals.

The human UGC angle: Content filmed by real creators doesn’t raise misleading conduct concerns under the ACL (assuming the content itself is truthful). There’s no AI-washing risk when the content is authentic, no synthetic voice disclosure requirement, and the only obligation is standard #Ad disclosure for paid partnerships.

South Korea: virtual person labels (in force since June 1, 2026)

South Korea announced the most ad-focused AI plan on this list in December 2025, and parts of it now apply. It’s worth separating what’s in force from what’s still planned, because many summaries treated the whole plan as law from January 2026.

What’s in force for advertisers

• Virtual person labels (since June 1, 2026). The Korea Fair Trade Commission’s revised endorsement guidelines now treat AI-generated virtual people as endorsers. Video ads must show a phrase such as “virtual person” near the figure while it’s on screen, and blog or community posts need a notice like “includes a virtual person” in the title or opening. A virtual person can’t claim to have used the product. Breaking these rules can count as unfair advertising under the Fair Labeling and Advertising Act, where fines go up to 2% of related revenue.
• AI Basic Act (since January 22, 2026). Providers of generative AI must notify users and label AI-generated output. These duties fall on AI providers, not advertisers, and fines (up to KRW 30 million) have a grace period of at least a year.
• AI fake experts in cosmetics ads (from November 27, 2026). An amendment to the Cosmetics Act bans ads where AI-generated doctors or experts endorse cosmetics.

What’s still pending

A Network Act amendment introduced on June 18, 2026 would require labels on all AI-generated content and ban removing them. Until it passes, there’s no blanket labeling rule for AI ads. The “five times damages” rule often quoted with the plan comes from a separate law on false and manipulated information (in force since July 7, 2026), not from an ad labeling law.

Imagine you’re a skincare brand running a TikTok campaign in Seoul with an AI-generated presenter. Today, she needs a “virtual person” label on screen whenever she appears, and she can’t say she uses your serum. From November 27, if she’s styled as a dermatologist, the ad isn’t allowed at all.

The human UGC angle: Real creators aren’t virtual people, so the label doesn’t apply to them. Standard paid partnership disclosure is still required.

China: the “dual-label rule” (effective September 2025)

China’s approach is the most technically prescriptive on this list. The Administrative Measures for the Labeling of AI-Generated Content took effect on September 1, 2025, issued by the Cyberspace Administration of China alongside a mandatory national standard.

The system requires two layers of labeling. Explicit labels are visible to users: text, icons, or audio cues clearly stating “AI-generated.” For images, the label text must be at least 5% of the shortest side of the image. Audio labels can use spoken indicators or even a Morse code rhythm encoding “AI.” Implicit labels are metadata embedded in the file itself, recording content attributes, the provider’s name, and a content identifier.

This dual requirement sits on top of China’s earlier Deep Synthesis Provisions (effective January 2023, which already required watermarking deep-synthesis outputs) and the amended Cybersecurity Law (effective January 2026, which adds a dedicated AI compliance provision covering ethics, risk monitoring, and safety assessment). The penalty structure for non-labeling includes warnings, public notices, rectification orders, possible suspension of information updates, and fines under related laws (the Measures don’t set fines of their own). It’s the same dual-layer approach platforms now apply worldwide, which we cover in our platform AI labeling guide.

Enforcement has started. In late April 2026, the Cyberspace Administration of China took its first public action under the labeling rules, summoning the operators of CapCut (Jianying), Maoxiang and Jimeng (Dreamina), issuing warnings and ordering fixes. No fines were reported. In June 2026, China’s market regulator and five other agencies also published non-binding guidance for the ad industry that singles out illegal AI-generated advertising.

The human UGC angle: Content created by real humans carries no labeling obligation under the AI content measures. No explicit or implicit labels required.

India: the “platform liability rule” (effective February 2026)

India’s IT (Intermediary Guidelines) Amendment Rules 2026 took effect on February 20, 2026. The rules target “synthetically generated information” (SGI), which covers any computer-generated or altered audio, visual, or audiovisual material that appears authentic.

The compliance burden falls primarily on platforms. Significant Social Media Intermediaries must require users to declare whether content is synthetically generated before publication, verify those declarations through automated tools, and label SGI prominently, with audio disclosures for audio SGI. Where feasible, platforms must also embed permanent metadata or unique identifiers to trace the source. In April 2026, MeitY published a draft amendment that would require labels to stay continuous and clearly visible; it hadn’t been finalised at the time of writing. On October 8, 2026, it also advised platforms to label manipulated or misleading synthetic content.

The enforcement mechanism is indirect but powerful. Non-compliant intermediaries lose their safe harbour protection from liability for third-party content, which means they can be held liable for the content itself. The rules also introduce a three-hour takedown deadline for government or court-ordered content removal, a 92% reduction from the previous 36-hour window. Deepfake intimate images and impersonation must come down within two hours.

ASCI’s rules for advertisers (from about December 2026). India’s advertising self-regulator, ASCI, published its Guidelines for Responsible Labelling of Synthetically Generated Content in Advertising on September 17, 2026. They apply three months after publication. Ads need a label for synthetic influencers and for material AI depictions, fabricated testimonials and non-consensual deepfakes are banned even with a label, and minor edits are exempt.

The human UGC angle: Content filmed by real creators isn’t synthetic under either set of rules. Minor AI edits, such as color correction, don’t need an ASCI label, but an AI-generated product result inside a creator video would.

Turkey: AI disclosure in ads (in force since August 1, 2026)

Turkey amended its Commercial Advertising and Unfair Commercial Practices Regulation in the Official Gazette on July 1, 2026, effective August 1. Advertisers must clearly disclose the use of AI or human-like digital characters when it significantly affects consumers’ buying decisions. And an AI replica of a real person can’t be shown using or recommending a product they never used or recommended.

Fines in 2026 range from TRY 99,339 to TRY 39,916,524, and they’re updated every year.

The human UGC angle: Real creators aren’t covered by the AI rule. Heavy AI editing that changes how a product looks or performs would still need a disclosure.

Brazil: the “pending legislation” wildcard

Brazil’s AI landscape is defined by a bill that’s close to becoming law but hasn’t crossed the finish line yet. PL 2338/2023 was approved by the Senate on December 10, 2024 and remitted to the Chamber of Deputies on March 17, 2025. It’s still in the Chamber’s special committee, where planned votes in May and June 2026 slipped. A vote is now expected after the October 2026 elections, and the bill then returns to the Senate.

The bill adopts an EU AI Act-inspired risk-based classification system, following the same tiered framework of prohibited, high-risk, and lower-risk AI systems. Proposed penalties are significant: up to R$50 million (~$8.5M) per violation or 2% of company revenue in Brazil, whichever is higher, alongside warnings, public disclosure of infractions, and suspension of AI system operations.

While the main AI bill is pending, Brazil’s ECA Digital (Law 15.211, enacted September 17, 2025 and in force since March 17, 2026) has already introduced platform obligations, including a ban on profiling-based ad targeting of minors, for protecting minors from harmful digital content, including AI-generated material. This signals that even without a comprehensive AI law, sector-specific rules are moving forward.

On the advertising side, Brazil’s self-regulator CONAR published a new influencer advertising guide on May 12, 2026 (applied from June 1), which covers AI-generated ads and virtual profiles.

The human UGC angle: No AI-specific advertising law is currently enforceable in Brazil. When PL 2338 passes, human-created content will sit outside the AI transparency and disclosure obligations.

Japan: the “soft law” exception

Japan sits at the opposite end of the regulatory spectrum from South Korea and China. The AI Promotion Act (effective June 2025) is the most permissive AI law passed by any major economy: no fines, no banned applications, no pre-launch approvals. Guidance under the Act (December 2025) encourages watermarking and provenance information where needed, without a binding labeling duty. The revised AI Basic Plan (July 14, 2026) commits to reviewing the law, possibly adding penalties. The broader approach relies on voluntary industry codes coordinated through the Hiroshima AI Process. Singapore takes a similar voluntary approach through its Model AI Governance Framework, with no binding ad rules or penalties.

Japan does enforce one rule that matters for creator content: stealth marketing has been banned since October 2023. On August 6, 2026, the Consumer Affairs Agency issued an order against SNOW Corp., which had paid influencers to post app-edited images without disclosing they were ads.

The human UGC angle: Japan imposes no AI-specific overhead yet, but the SNOW case shows that paid creator posts still need a clear ad disclosure.

The compliance pattern across markets

Three regulatory tiers have emerged, and understanding which tier a market falls into shapes how you build your compliance program.

Hard law: China (dual labels), India (platform liability, plus ASCI’s ad labeling rules), Turkey (AI disclosure in ads) and South Korea (virtual person labels, with broader labeling pending). If you’re running AI-generated ads in these markets, labeling isn’t optional.

Consumer law plus self-regulation: the UK (fines up to 10% of global turnover), Canada (Ad Standards guidelines backed by the Competition Act) and Australia (up to A$100 million). No AI-specific statutes, but regulators are actively enforcing. Don’t mistake the lack of a dedicated AI law for a lack of enforcement risk.

Soft law: Japan (voluntary framework, no penalties) and Brazil (bill pending). Building compliance now puts you ahead of wherever these markets land next.

Every jurisdiction on this list is moving toward requiring disclosure of AI-generated content in advertising. The only variables are speed and enforcement mechanism. Brands that build compliance around the strictest current standards position themselves for whatever comes next in every other market.

The bigger picture

Every country on this list is converging on the same expectation: if your ad creative is AI-generated, disclose it. The mechanisms differ, but the direction is uniform. And it’s accelerating.

For brands running AI-generated creative across multiple markets, the compliance surface is getting wider with each new rule. For brands working with real human creators, it stays close to the same everywhere: disclose the paid partnership, and add an AI label only where AI changes what viewers see.

Here’s a full UGC AI compliance playbook that walks through the practical steps: how to audit your creative pipeline, build labeling workflows for hard-law markets, update creator briefs, and decide where human UGC simplifies your operations.

Looking for proven human creators and AI analytics? Check out Billo CreativeOps.

FAQs