A Survey, Not a Statute
On May 21, 2026, South Korea's Korea Fair Trade Commission (KFTC) opened written inquiries under Article 87(1) of the Monopoly Regulation and Fair Trade Act against 46 firms — 29 AI service developers and 17 companies embedding AI into their products, a group that includes Samsung Electronics and Apple. The agency calls it the 'AI Service Market Survey' and frames it as the third leg of a multi-year inquiry that began in 2024 with foundation models and AI chips, and continued in 2025 with data-handling practices across seven online sectors. A policy report titled 'AI Downstream Markets and Competition' is promised before year-end, with a consumer-side survey following in July.
The instinct in some quarters will be to read this as the start of another digital-markets regime in the EU mold. It is not — at least not yet. A market study under Article 87 is information-gathering, not enforcement. The KFTC has no penalty power flowing directly from the survey, no designation list, no mandatory non-discrimination duties, and no advance prohibitions. What it does have is a methodology that has, twice before, produced reports rather than rules.
Steelmanning the Probe
There is a serious case for the survey. South Korea has lived through a year of consumer-AI controversies that look exactly like the harms competition agencies are meant to police. The Seoul YMCA filed a complaint with the KFTC in March 2025 alleging that Apple overstated the readiness of Apple Intelligence features on Korean iPhones, and the regulator has spent more than a year requesting data without acting. Apple has agreed to compensate U.S. consumers for the same conduct; on present trends, Korean buyers will not see equivalent relief. If 'false or exaggerated advertising' of AI features is a live consumer-protection issue, mapping who controls the consumer-facing surface — the phones, cars and social apps where AI now lives — is a reasonable place for an antitrust agency to start.
Market concentration at the integration layer is also a non-frivolous concern. Naver holds roughly 52.7% of Korean search and Baemin around 64.9% of food delivery, and those are the firms most likely to bolt generative features onto already-dominant interfaces. If competitive harm emerges in Korean AI, it will more plausibly emerge at the device-and-app edge than at the model layer, where global frontier developers compete fiercely on price and capability. The KFTC's December 17, 2024 report on generative AI already flagged the structural concerns upstream — NVIDIA's grip on accelerators, AWS's cloud share, expertise-poaching by incumbents. Extending the inquiry downstream is internally consistent.
Where the Calibration Sits
The right question is not whether to look, but what to do with what is found. Three things suggest the KFTC is, for now, picking the lighter-touch instrument.
First, it chose a market study over a fresh ex-ante bill. The Online Platform Regulation Act — Korea's nearest analogue to the EU Digital Markets Act — has been stalled for two years, and the KFTC itself pivoted in September 2024 toward amending existing competition law rather than building a new designation regime. The current Lee administration's antitrust chief, Ju Biung-ghi, told an industry audience in December 2025 that 'in digital markets, time matters,' and signaled a preference for consent resolutions over multi-year litigation. A market study is the upstream of that approach: identify, then act case by case.
Second, the survey explicitly leaves room for the startup story. The KFTC's own statement said the goal is 'an environment in which domestic startups can compete fairly.' That language matters because Korean venture associations have warned that DMA-style ex-ante duties could discourage venture investment in platform startups and deter foreign capital. A targeted inquiry that surfaces specific anticompetitive conduct — bundling, self-preferencing, exclusionary deals — does not chill startup formation; a list of pre-emptive obligations on every firm above a revenue threshold does.
Third, the consumer-protection angle has a narrower, better-fitting tool than competition law. AI-driven misleading advertising is already addressed under Korea's Act on Fair Labeling and Advertising. CSIS has warned, correctly, that DMA-style structural remedies tend to shift rents to intermediaries rather than benefit consumers. If the KFTC's downstream report becomes the basis for clearer labeling and disclosure standards, that is a proportionate outcome. If instead it becomes the gateway to ex-ante duties on Samsung phones and Apple devices because they integrate AI, it will repeat Europe's mistake of regulating market structure as a proxy for consumer harm.
What to Watch
Three signals will determine whether the AI Service Market Survey lands well. The first is whether the year-end 'AI Downstream Markets and Competition' report names specific conduct or specific firms; concrete findings beat vague structural worries. The second is whether the KFTC carries the Apple Intelligence false-advertising complaint to a conclusion under existing law, which would demonstrate that current tools work without a new designation regime. The third is whether the report feeds into measured amendments to the Monopoly Regulation and Fair Trade Act, or revives the Online Platform Act as an ex-ante regime — a path that has already drawn warnings from USTR Jamieson Greer and would expose Korean exports to retaliatory pressure.
South Korea's competition agency has, twice, produced AI reports without rushing into rules. The third installment, focused on the consumer edge where harm is most concrete, can be the most useful of the three — if Seoul resists the temptation to convert a fact-finding exercise into a presumption that bigness equals abuse.