Late summer is when skincare routines get heavier. The lightweight gel that worked in August gets swapped for a cream, and the ingredient list on the back of the new jar is longer and less familiar than the old one. Ceramides, squalane, petrolatum, dimethicone, urea, a few silicones with numbers in their names.
Several of those names carry a bad reputation online. Petrolatum is a petroleum derivative. Silicones are said to suffocate skin. Mineral oil turns up on avoid-lists that circulate every year. None of those claims come from regulators, but they shape what people buy.
So we checked what regulators actually say. We queried our database of 21,796 cosmetic ingredients and 31,191 regulatory entries across 10 markets, and pulled every restriction attached to the ingredients that make up a typical autumn moisturizer. Of the 26 we checked, 20 have no regulatory entry in any of the 10 markets. The six that do are mostly capped at levels above what formulators use anyway.
The 20 with no entries
These returned zero regulatory records across all 10 markets in our data.
| Category | Ingredients |
|---|---|
| Humectants | Sodium Hyaluronate, Hyaluronic Acid, Butylene Glycol, Propylene Glycol, Betaine, Sodium PCA, Trehalose |
| Barrier lipids | Ceramide NP, Ceramide AP, Ceramide EOP, Cholesterol, Squalane |
| Emollients | Butyrospermum Parkii (Shea) Butter, Dimethicone, Cyclopentasiloxane, Cyclohexasiloxane, Mineral Oil / Paraffinum Liquidum, Lanolin |
| Soothing actives | Madecassoside, Asiaticoside |
Propylene glycol, mineral oil and the silicones on that list are the ones most often flagged in consumer discussion. In our regulatory data, none of them carries an entry in any of the 10 markets we track. Neither does shea butter, which nobody worries about. Reputation and regulation are running on different tracks here.
The six that do carry restrictions
| Ingredient | Markets | Type | What the entry says |
|---|---|---|---|
| Petrolatum | EU, China, ASEAN, Brazil, Argentina | Prohibited | Exempt when the full refining history is known and the source material can be shown not to be a carcinogen |
| Urea | Korea, Taiwan, Canada | Restricted | Korea 10 percent. Taiwan 10 percent for hair dye products, 5 percent for other products. Canada 10 percent for cosmetics other than bath products |
| Panthenol | Japan | Restricted | Listed as D-/DL-Pantothenyl Alcohol. No cap on rinse-off products |
| Allantoin | Taiwan, Japan | Restricted | Japan sets 0.50 g/100 g for leave-on, excluding mucous membrane products |
| Bisabolol | Japan | Restricted | 1.2 g/100 g |
| Glycerin | Canada | Restricted | Manufacturers must verify the grade of the raw material for oral and leave-on products |
Read the conditions and most of these stop looking like barriers. Panthenol's Japanese entry explicitly exempts rinse-off products. Glycerin's Canadian entry is a raw material sourcing obligation on the manufacturer, not a limit on the finished product. Allantoin and bisabolol carry caps in Japan, but at 0.50 and 1.2 g per 100 g they sit above the levels these ingredients are typically used at.
Two entries do constrain formulation. Urea is capped in three markets, and the tightest of them is Taiwan's 5 percent for products other than hair dye, which is below the 10 percent that Korea and Canada allow. Anyone building a keratolytic body cream is working against that ceiling rather than the Korean one. And petrolatum is prohibited in five markets, though the prohibition carries an exemption written directly into the entry: material whose full refining history is documented, and whose source can be shown not to be a carcinogen, falls outside the ban. Cosmetic-grade white petrolatum meets that condition, which is why petrolatum balms sell in the EU without difficulty.
Canada's urea entry has a gap in it. It names two product categories, cosmetics intended for dilution in bath water and everything else, but carries a concentration figure only for the second. The first has no value in the source text. We have left it as it appears rather than filling it in.
The silicone case: regulation targets molecules, not families
Three ingredients sit next to each other in our records, at codes 768, 769 and 770.
| Ingredient | Common name | CAS | Ring size | Regulatory entries |
|---|---|---|---|---|
| Cyclotetrasiloxane | D4 | 293-51-6, 556-67-2 | 4 | Prohibited in EU and ASEAN |
| Cyclopentasiloxane | D5 | 541-02-6 | 5 | None |
| Cyclohexasiloxane | D6 | 540-97-6 | 6 | None |
The three are the same cyclic dimethylpolysiloxane structure with a different number of units in the ring. Our ingredient records define each one that way explicitly: D4 is the structure where n averages 4, D5 where n is 5, D6 where n is 6, each specified to contain under 1 percent of the other ring sizes.
Their stated functions are identical. All three are listed as hair conditioning agents, skin conditioning agents in the emollient subcategory, and solvents. The function strings match character for character across all three records.
D4 is prohibited in the EU and ASEAN entries in our data, under the name Octamethylcyclotetrasiloxane. D5 and D6 carry nothing. The ingredient-level classification field agrees, marking D4 as Prohibited and the other two as Not Listed, which gives us the same answer through a second path in the data.
Whatever the reasoning behind that line, its structure is what matters when reading an ingredient list. Regulators evaluate specific substances against specific evidence, one substance at a time. A rule written for D4 does not extend to D5 because the two share a chemical family and a function. "Contains silicones" tells you almost nothing about regulatory status. "Contains cyclotetrasiloxane" tells you something specific.
Where restriction actually concentrates
The 20-of-26 result is not really a fact about moisturizers. It is a fact about what regulators choose to write rules about.
Across our whole ingredient set, 862 of 21,796 ingredients carry at least one regulatory entry, or 3.95 percent. Grouping those 862 by their stated function, and calculating what share of each function group carries an entry, produces a spread of two orders of magnitude.
| Function | Ingredients with the function | Share carrying a regulatory entry |
|---|---|---|
| Hair dye | 71 | 88.73% |
| Preservative | 127 | 84.25% |
| Colorant | 102 | 72.55% |
| Denaturant | 33 | 48.48% |
| Anti-discoloration agent | 63 | 33.33% |
| Reducing agent | 42 | 30.95% |
| Fragrance | 1,333 | 12.53% |
| Skin conditioning, occlusive | 596 | 1.01% |
| Skin conditioning, emollient | 2,223 | 0.63% |
| Skin conditioning, humectant | 1,506 | 0.53% |
The top of that table is made up of ingredients designed to cause a reaction. Hair dyes oxidize. Preservatives kill microbes. Colorants carry pigments and, historically, heavy metals. The bottom is made up of ingredients that hold water or form a film, which is physical rather than biological work. Hair dye sits at 88.73 percent. Humectants sit at 0.53 percent.
Counting by raw numbers instead of rates inverts the result: fragrance ingredients contribute 167 of the 862 restricted ingredients, more than any other function, but that is because there are 1,333 fragrance ingredients to begin with, and the rate is 12.53 percent. And a high rate is not a danger rating. Preservatives sit at 84.25 percent largely because most markets regulate them through positive lists with concentration caps, so appearing on the approved list is itself recorded as a regulatory entry. A low rate, equally, is not a safety clearance.
What "no entry" does and does not mean
A negative result is easy to over-read, so the limits of this one are worth setting out.
Absence in our data is not permission. A zero means we hold no regulatory entry for that ingredient in that market. Regulatory systems differ in structure. Some publish exhaustive prohibited lists, others regulate through general safety obligations that never name a specific ingredient. The US carries 111 entries in our data in total, against 5,342 for the EU, so a zero in the US column carries almost no information.
Some rules attach to the source, not the substance. Lanolin returns no entry against its ingredient record, but it appears inside a Taiwanese prohibition covering wool grease derived from countries affected by BSE. The rule targets the origin of the raw material rather than the ingredient itself, so it does not link to the ingredient record. Rules of this shape are invisible to an ingredient-level query, and lanolin is unlikely to be the only case.
We counted markets, not rows. Petrolatum returns 10 rows in our data but covers 5 markets, and D4 returns 4 rows covering 2. In both cases two source ingredient codes in the Korean MFDS data carry identical restriction text, producing duplicate rows per market. We compared the petrolatum pairs field by field and they match character for character, differing only in record identifiers that sit one apart, which suggests adjacent entries in the original notice. We count 664 excess rows of this kind across the database. Anyone citing these figures should count distinct markets rather than records.
Brazil and Argentina are not independent evidence. Both return 4,052 entries in our data, identical down to the split between prohibitions and restrictions. Argentina's cosmetic ingredient rules follow the Mercosur harmonized list, which Brazil also applies. We count them as two markets because they are two regulatory authorities, but a restriction appearing in both is one decision, not two.
Methodology and sources
Ingredient data comes from the Korea Cosmetic Association ingredient dictionary. Regulatory entries are compiled from Korea's Ministry of Food and Drug Safety open data covering restricted cosmetic ingredients, and from EU CosIng Annexes. Figures reflect our database as of August 2026 and cover 10 markets: the EU, Korea, ASEAN, China, Japan, Taiwan, Brazil, Argentina, Canada and the US.
Every figure in this post came from four queries against that dataset: an ingredient lookup by INCI name, a regulations lookup by ingredient code, a count of entries grouped by market, and a count grouped by stated function. The first two are the queries anyone would run to check a single ingredient against 10 markets.
Matching was done on ingredient records rather than by searching restriction text. A free-text search for "Glycerin" across restriction text returns seven hits, all of them nitroglycerin. A search for "Hyaluronic Acid" returns two, both a gold nanoparticle compound. Substring matching on ingredient names produces false positives at a rate that would make this entire analysis wrong.
Regulatory data changes, and the absence of an entry should be verified against the current official source before it is relied on for a compliance decision.
For the underlying data, see K-Beauty Cosmetic Ingredients on RapidAPI.
Comments
Post a Comment