Korea Is Making Cosmetic Safety Assessments Mandatory. We Counted What That Asks of an 87-Ingredient Product.

The longest ingredient list we sell runs to 87 entries. It belongs to a spicule ampoule that ships as two 1.4g vials, and reading the label end to end, out loud, takes about four minutes.

From 2028, somebody has to write a safety file for every one of those 87 lines. Not for the product. For each ingredient in it.

Korea amended the Cosmetics Act at the end of 2025 to make safety assessment mandatory, and in July 2026 the Ministry of Food and Drug Safety put the implementing details out for comment. The comment window closed yesterday. We wanted to read the draft and count what it would actually demand of our worst-case product, so we went to the MFDS site to pull the notice.

What we could and couldn't read

The notice went out on 8 July 2026 and comments closed on 18 August. That much is reported consistently. Past that we're relying on secondary coverage, and the coverage doesn't fully agree with itself.

Three Korean trade outlets put the company-size threshold at annual production or import of 1 billion won. A fourth rendered the same figure as 10 billion. One of those is a unit-conversion slip and we're fairly confident which, but we haven't seen the notice, so we're telling you the number is 1 billion won and that we couldn't verify it at the source.

The phase-in, as reported: functional cosmetics in 2028, infant and children's products in 2029, newly manufactured or imported products in 2030, everything in 2031. Companies under the threshold skip the middle steps and land on children's products in 2029 and everything in 2031.

Note what that sequencing implies. The categories arrive in order of how much regulatory attention they already get, which means the products carrying the strongest existing claims are asked to prove themselves first, and the ordinary moisturiser at the back of the queue gets three extra years. Whether that is sensible triage or simply the path of least resistance depends on whether you think claim strength tracks actual risk. Our own data says it doesn't especially: the eight worst ingredient grades we published this year were spread across ordinary shampoos and foot creams, not functional cosmetics.

So the honest version of "Korea is making safety assessments mandatory" is that a small brand launching a face cream this month may not owe a full ingredient dossier until 2031. Five years is not nothing.

Eighty-seven ingredients, one file

Our catalogue is 54 products. The median ingredient list is 33 entries long and the mean is 34.2, which sounds tidy until you look at the ends. The shortest list we sell has exactly one entry, a jojoba oil that is nothing but Simmondsia Chinensis seed oil. The longest has 87.

That's an 87-fold spread inside one small shop.

The EU's Annex I, Part A, is the readable statement of what a safety file holds. Per product it wants quantitative and qualitative composition with each ingredient's function, physicochemical characteristics and stability, microbiological quality, impurities and traces and packaging interactions, normal and reasonably foreseeable use, exposure to the product, exposure to the substances, a toxicological profile for each substance, and undesirable-effects data. Part B then wants a conclusion, warnings, the reasoning, and a named assessor's signature.

Read that list again with 87 in mind. Most of those items are written once per product: one stability statement, one microbiological result, one packaging assessment. They don't care whether the formula holds nine ingredients or ninety.

One item does care. "Toxicological profile of the substances" is singular per substance, and our ampoule has 87 of them.

That's the whole asymmetry. A short formula and a long formula cost about the same to document everywhere except the one place where the long formula costs nine times more. Which is a strange incentive to build into a rule, because ingredient count and ingredient risk are not the same thing, and our own data says so: the 87-line ampoule carries no high-risk flags, while our worst grades this year sat in products with fewer than half that many entries.

Korea has been doing this for babies since 2019

The 2026 move is being covered as a new regime. It reads more like an expansion of one.

Article 4-2 of the Cosmetics Act has, since a 2019 amendment, required makers of infant and children's cosmetics to prepare and keep three sets of documents: a product and manufacturing description, safety assessment data, and efficacy substantiation. MFDS later audited the top 36 responsible sellers, roughly 80% of that market, and found three managing the paperwork inadequately.

Which means the machinery exists, the guideline format exists, and the enforcement precedent exists.

What changes in 2028 is the denominator.

The EU has had the readable version since 2013

Article 10 of Regulation 1223/2009 says a cosmetic can't go on the EU market until a responsible person has ensured a safety assessment was done and a safety report built per Annex I. It also says who may sign: someone holding a university degree in pharmacy, toxicology, medicine or a similar discipline. Non-clinical studies behind it must follow Good Laboratory Practice.

The arithmetic sits in the SCCS Notes of Guidance, currently the 12th revision, adopted in December 2023. The core figure is the Margin of Safety, which divides a no-observed-adverse-effect level by systemic exposure dose, where exposure folds in concentration, how much product goes on per day, dermal absorption and body weight. A margin of 100 is the conventional floor.

Now the part the alarmed coverage tends to skip. The Korean guideline issued in June 2026 reportedly exempts ingredients already governed by fixed legal limits from the MoS step, which takes preservatives, colourants and UV filters off the calculation pile. It also reportedly lets companies file concentration ranges rather than exact percentages.

Our 87-ingredient ampoule carries 1,2-hexanediol and a UV-irrelevant set of extracts. Strip the fixed-limit categories out of an average list of ours and the calculation count drops by a meaningful chunk, though we can't give you the exact number without the final rule.

We also went looking for a per-ingredient price for a dossier and gave up. Nobody publishes it. Consultancies quote per product, those quotes assume a normal 25-ingredient formula, and not one of the pages we could open would put a figure against an 87-line list.

So we can't tell you what this costs. We can tell you what it requires, which is a qualified named human who signs their own name to a conclusion. That person is the scarce input here, not the toxicology.

What our screening found

The practical takeaway

Nothing here changes what you should buy this year. It changes what a label will mean in about five, and it changes who is accountable when it's wrong.

Right now, if an ingredient in your moisturiser turns out to be a problem, the paper trail behind that decision mostly doesn't exist outside the EU. After 2031 in Korea, it will, and it will have a name on it.

If a brand tells you in 2029 that its products are "safety assessed," ask which category and which year, because until 2031 that sentence can be true of one SKU and untrue of the next one on the same shelf. Functional cosmetics come first. Everything else trails.

And if you want the version of this that's already enforceable, look at anything sold into the EU, where the file has been mandatory since 2013 and a named assessor has been on the hook the whole time. That's the same reason we publish the limits of reading a position on a label: the ingredient order tells you less than people think, and the document behind it tells you more than you'll ever see. Our scoring criteria sit in the same gap. So does our writeup on what spicules actually do, which is the category this 87-line ampoule belongs to.