MyAIDevelopment · Product · Operations notes

Law · Ethics · Ad Disclosure

This document is not legal advice. It is a map to help developers and POs notice "this needs checking". Confirm actual application through legal review and the latest original texts.

Three Questions

flowchart TD
    Q[Marketing Activity] --> D{Personal Data?}
    Q --> A{Is it an Ad?}
    Q --> G{AI Generated?}
    D -->|Yes| C[Consent and Purpose]
    A -->|Yes| L[Ad Disclosure]
    G -->|Yes| T[AI Labeling]
    C --> R[Legal Review]
    L --> R
    T --> R
  1. Do you use personal data? → Consent for collection and use, purpose limitation, consent for targeted advertising
  2. Is it an ad? → Ad labeling, disclosure of material connections, opt-in consent to receive, truthfulness and substantiation of claims
  3. Is it AI-generated? → Labeling of AI-generated content and virtual people

If you are building checkout, cancellation, or pricing screens, add one more: does the screen design distort the consumer's choice? → Korea's E-Commerce Act dark-pattern rules.

Korea: Sending Advertising Information (Network Act)

Article 50 of Korea's Act on Promotion of Information and Communications Network Utilization and Information Protection (Network Act) and its Enforcement Decree set the rules for sending commercial advertising information through electronic transmission media (text messages, email, app push, etc.).

Requirement Content
Prior opt-in consent In principle, the recipient's explicit prior consent is required
Night-time sending From 9 p.m. to 8 a.m. the next day, separate prior consent is required
Labeling duty An "(Ad)" label where the advertising information begins, the sender's name and contact, and opt-out instructions
Refusal and withdrawal No sending after the recipient refuses or withdraws consent
Consent reconfirmation Reconfirm consent every two years from the date it was given (Enforcement Decree Article 62-3)

Bad example:

Bundle marketing consent into the mandatory signup terms and send an event push at 10 p.m.

Good example:

Separate marketing consent as an optional item and send night-time messages only to those who gave separate consent. Put "(Ad)" at the start of every message and include an opt-out path.

Korea: Personal Data and Targeted Advertising (PIPA)

Korea: Labeling and Advertising Act (Unfair Labeling and Advertising)

Article 3(1) of Korea's Act on Fair Labeling and Advertising prohibits four kinds of labeling or advertising that are likely to deceive or mislead consumers and to harm fair trade.

Type Meaning Easy-to-trip claims for the example product
False or exaggerated Stating things contrary to fact, or overstating facts Advertising an integration still in beta as "fully supported"
Deceptive Concealing or downplaying facts Advertising a "free plan" while hiding its 7-day retention limit
Unfairly comparative Comparing without stating what is compared and on what basis, or claiming superiority without objective grounds "3x faster than competitors" with no conditions
Disparaging Belittling another business or product without objective grounds, or advertising only unfavorable facts about it "Company A's tool goes dark during incidents"

Article 5 sets a duty to substantiate. A business must be able to substantiate the factual claims in its labeling and advertising, and must submit substantiating material within 15 days when the KFTC requests it. In practice, to say "first", "No. 1", or "N times", you need the evidence before the ad runs.

Bad example:

Korea's first and industry-leading incident analysis tool

Good example:

In-house benchmark as of 2026-06: with 100,000 log lines and 20 deploys, finding candidate causes was 1.8x faster than manual searching. (Method published)

The number in the good example is an assumption for illustration. The point is to state the date, conditions, comparison baseline, and measurement method together and to keep the evidence. AI-generated ad copy and images are held to the same standard.

Korea: E-Commerce Act Dark-Pattern Rules (in Force Since 2025-02-14)

With the amended E-Commerce Act taking effect on 2025-02-14, six types of online dark patterns became subject to statutory regulation. The KFTC published a Q&A just before the rules took effect, and an amended Enforcement Decree set fines and business-suspension criteria for violations. The rules cover online transactions with consumers; a SaaS that individual developers pay for by card may be covered, so confirm with legal review.

Type Description Basis
Hidden renewal Raising a subscription price or converting free to paid without the consumer consent and notice procedure Act Article 13(6), Enforcement Decree
Drip pricing Showing only part of the price at first and revealing extra charges during checkout Act Article 21-2(1)
Preselected options Preselecting options that favor the business so the consumer pays with them Act Article 21-2(1)
False hierarchy Making one choice stand out by size, color, or position so consumers are misled Act Article 21-2(1)
Obstructing cancellation or withdrawal Cancellation or account deletion harder than signup; restricted ways to cancel Act Article 21-2(1)
Nagging Repeatedly asking, via pop-ups and the like, to change a choice already made; excepted if the consumer can opt not to be asked again for at least 7 days Act Article 21-2(1), Enforcement Decree

Rule for paid conversions and price increases: when a subscription price increases or a free service converts to paid, the business must obtain the consumer's consent within the 30 days before the increase or conversion, and must tell the consumer the conditions and methods for withdrawing that consent (Enforcement Decree).

Bad example:

When the 14-day free trial ends, charge the card automatically without notice, and accept cancellations only by email to support.

Good example:

Within the 30 days before conversion, tell the user the conversion date, amount, and how to cancel, and get explicit consent. Without consent, the account stays on the free plan. Cancellation happens in the same settings screen as signup, in the same number of steps.

Korea: Endorsements and AI Virtual People (Korea Fair Trade Commission)

Korea: AI Basic Act

In marketing practice, split the cases like this.

Situation Rules to check first
You make ad images, copy, or virtual models with AI tools (a user) Virtual-person disclosure under the KFTC endorsement guidelines; the Labeling and Advertising Act's ban on unfair ads
Your product provides a generative AI feature (e.g., it shows customers incident-cause summaries written by generative AI) As an AI business operator, advance notice and output labeling under AI Basic Act Article 31

Even when you are a user with no AI Basic Act duty, AI output used in ads still carries the disclosure duties of other laws.

EU

United States

Practical Checklist

Ethics: Lines to Hold Before the Law

References