The Korea Fair Trade Commission is taking comment to 11 November 2026 on a draft AI advertising guideline that targets AI washing

Key takeaways

• The draft is for comment only. The KFTC says it will finalise it after comments and consultation with the ministries concerned.
• It covers four types of improper advertising: false or exaggerated, deceptive, improper comparison and disparaging.
• It includes 16 examples and an 8-item self-check list. Not meeting the list is not itself a violation, the release summary says.
• Comments may be sent by post or email until 11 November 2026.

Is the KFTC draft guideline binding, and what is open for comment?

No. The Korea Fair Trade Commission (KFTC) is collecting comment on a draft titled "인공지능(AI) 관련 표시·광고 가이드라인(안)" (Guideline on AI-related Labelling and Advertising). The KFTC release says the draft has no legal binding force but is meant to make the Fair Labelling and Advertising Act more predictable to apply and to prevent violations in advance: 법적 구속력은 없으나 표시광고법 적용의 예측가능성을 높이고 (release, page 2). The draft itself carries a notice that it is a draft for comment and not a final text: 본 자료는 의견을 청취하기 위한 초안으로 확정안이 아님을 유의하시기 바랍니다 (draft, page 1).

A footnote in the release adds that the guideline does not newly impose stronger, separate duties on AI advertising than on ordinary goods. It explains the Act's general principles of truthfulness, completeness, fairness and substantiability using AI-specific examples: 인공지능 관련 표시·광고에 일반 상품보다 강화된 별도의 의무를 새로 부과하는 것이 아니라 (release, page 2, footnote). The release says the KFTC will finalise and publish the guideline after the comment results and consultation with the ministries concerned, and plans to supplement it periodically as decisions accumulate and the technology changes.

What does the draft say about AI washing and the four types of improper advertising?

The release says AI is often used in wording that suggests advanced, accurate and objective products, and that consumers may expect better performance or even pay more. It cites a Korea Consumer Agency survey from November 2025 in which 57.9 percent of respondents said they would buy an AI product even if it cost more than an ordinary one, at an average of about 21 percent extra.

Against that, the release names two problems. One is claiming AI use that does not exist ("AI washing") or using AI but not disclosing important matters such as its scope of application, operating conditions or performance limits. The other is false or exaggerated advertising that uses AI-generated virtual people, reviews and deepfakes, which the release says spreads through social media mainly in food and pharmaceutical categories and harms consumers including older people.

The draft sorts improper AI advertising into the four types the Act prohibits: false or exaggerated, deceptive, improper comparison and disparaging. It gives 16 item-level examples, which the release describes as hypothetical cases reconstructed from real consultations and investigations, across electrical and household goods, food, cosmetics and household chemicals, and education, legal and financial services. The summary table counts seven false or exaggerated examples (including an AI lawyer), five deceptive (including an unlabelled AI virtual-expert recommendation and a deepfake pseudo investment-advice case), two on comparison and two on disparagement.

What do the self-check list and the AI-generated content guidance cover?

The draft includes a business self-check list of 8 items in four steps: whether AI is actually used, where and under what conditions, whether the effect is described specifically, and whether it can be objectively proven. Items include not presenting third-party AI as self-developed and giving information that affects performance, such as the range of training data. The release summary says that not meeting the list is not in itself a violation or a ground for a penalty.

The summary also reports that the draft distinguishes itself from the transparency duty in Article 31 of the AI Basic Act (labelling of AI-generated output), which is a separate instrument. On AI-generated images, video and reviews, the summary says they may be used in advertising but that the use of AI should be clearly shown where consumers could take them for real people, products or results, and that AI-generated reviews should not be presented as real consumer experience. It also reports that AI virtual-expert recommendations in food, cosmetics and drug advertising would be banned under revised sectoral laws taking effect on 27 November 2026, which we did not review. The draft also contains consumer notes and FAQs (10 for businesses and 6 for consumers).

What should marketers and counsel do now?

Because this is a non-binding draft, nothing here creates a compliance deadline. Companies that advertise AI features in Korea, and foreign brands selling into it, may want to decide whether to file comments by 11 November 2026, since the KFTC has asked for views for, against or proposing amendments, with reasons.

• Compare each AI claim in Korean-market advertising with the draft's four improvement points: where AI is applied, what performance is claimed, the conditions needed for it to work, and the stated limits.
• Check that any claim of AI use reflects what the product actually does, and that third-party AI is not described as in-house.
• Keep test results or other substantiation for performance claims. The release summary says the burden of proving factual claims lies with the business.
• Flag AI-generated personas, reviews and video for the draft's disclosure guidance and read the full draft before relying on this summary.

What are the key dates and parties?

What we did not verify

What we opened: the KFTC press release as reproduced on the Korean government policy portal (korea.kr), its attached press-reference PDF including the summary, and the attached 60-page draft guideline PDF, on 24 October 2026.

What we did not open: the KFTC's own website page, the Fair Labelling and Advertising Act, the AI Basic Act, the sectoral-law amendments said to take effect on 27 November 2026, the MSIT notice on the AI confirmation scheme, or the 2026 revision of the recommendation and endorsement review guideline. We read the draft only for its cover notice and did not review its 16 examples line by line.

What we refuse to claim: that the guideline is final or binding, that any business has been found in breach, or what the KFTC will do after comments close. Facts that appear only in the release summary are attributed to it.

Related briefings

• AI Regulation News hub
• Asia-Pacific AI regulation tracker
• AI disclosure rules
• Korea KFTC notice on substantiating AI safety claims
• South Korea MSIT notice on AI product and service confirmation
• South Korea AI Act generative content labeling
• UK ASA monitoring of AI advertising disclosure