EU CosIng — European Commission Cosmetic Ingredients Database

Every cosmetic product sold in the EU has its ingredients printed in a specific vocabulary, and somebody has to own that vocabulary. That somebody is the European Commission's Directorate-General for Internal Market, Industry, Entrepreneurship and SMEs (DG GROW), and the register is CosIng: the official information database covering cosmetic ingredients under Cosmetics Regulation (EC) No 1223/2009, the Inventory and Common Nomenclature of Cosmetic Ingredients amended by Decision 2006/257/EC, and the opinions of the Scientific Committee on Consumer Safety.

The database holds two distinct kinds of record, and the distinction matters more than anything else about it. Ingredients are the inventory entries - the common nomenclature written in capital letters that EU labelling uses, things like MARIS AQUA. Substances are the regulated chemical compounds that correspond to entries in the Regulation's annexes: Annex II prohibited substances, Annex III restricted, Annex IV colorants, Annex V preservatives, Annex VI UV filters, plus CMR references under Annex 15. One record type exists so labels can be read consistently; the other exists because the law says so.

Each record carries a dense identity card: INCI name, US INCI variant, INN/ISO/AN and Ph. Eur. names, chemical and IUPAC descriptions, CAS number, EINECS/ELINCS/EC number, colour index number, assigned cosmetic functions from the Commission's controlled list, annex number and restriction wording, maximum concentration limits, SCCS opinion references, perfuming flags, Active/Inactive status and Official Journal publication markers. Content traces back to the 1976 Cosmetics Directive, and the corpus runs to roughly mid-tens-of-thousands of ingredient and substance entries.

One boundary condition shapes how to use it: the Commission states plainly that only information provided by the Regulation itself has legal value, that cross-references between substances and ingredients are not binding, and that inclusion in the inventory does not turn it into a positive list of permitted substances. Get a sample of this dataset and Datadory hands you the resolved personal-care slice - records with their fields populated and their annex context intact - rather than a pointer into someone else's search box.

Three records verified during the August 2026 research pass, shown flat exactly as the database describes them:

inci_name : MARIS AQUA
us_inci   : SEA WATER
record    : ingredient
functions : HUMECTANT, SOLVENT
chemical  : Sea Water is water obtained from the sea or from inland bodies of salt water
cas/ec    : - / -
status    : Active

inci_name : POTASSIUM ALUMINUM POLYACRYLATE
functions : ABSORBENT, VISCOSITY CONTROLLING
chemical  : 2-Propenoic acid, homopolymer, aluminum potassium salt
cas       : 67785-56-2
ec        : -
status    : Active

inci_name : PEG-20 COCAMIDE
functions : (none recorded)
cas       : 68425-44-5
ec        : (blank)
annex     : none attached

Read them as three different grades of evidence. Row one is a pure inventory entry - sea water under its EU name and its US variant, dual-hatted as humectant and solvent, with no CAS number because there is nothing to register. Row two is the chemically specific case: a polyacrylate salt carrying both a CAS registry number and a proper IUPAC-style description, doing two jobs in a formula at once. Row three is the honest end of the corpus - PEG-20 COCAMIDE is a real, shippable record that nonetheless carries no functions and no annex attachment, which is precisely why blank-field handling matters when you pipe this into a compliance tool.

Substance records extend the same shape with the legal layer: an annex_no placing the compound in II through VI, the restriction wording stating conditions and warnings, a max_concentration ceiling for finished products, and the SCCS opinion reference that justifies the treatment. That annex layer is the part of the record a regulatory team actually reads.

Twelve documented fields carry each record, and they sort into three jobs. Identification fields - inciName, inciUsaName, casNo, ecNo, substanceId - pin down exactly which molecule a label word refers to, and the CAS and EC numbers do double duty as join keys into chemical-supplier and safety-data universes. Classification fields - itemType, annexNo, functionName, chemicalDescription - say what the thing is and what job it does in a formulation, with the controlled function list turning free-text chemistry into queryable categories. Compliance fields - cosmeticRestriction, maximumConcentration, status - carry the legal weight: what conditions apply, what ceiling applies, and whether the entry is live.

Definitions below were checked against live responses during the August 2026 research pass, so the table describes what arrives rather than what a brochure promises. Substance records can also carry SCCS opinion references, Official Journal publication flags, colour index numbers and INN/ISO/AN/Ph. Eur. alternative names - these fold under additional fields on request so the core schema stays honest about what ships first.

Geography - the database covers the European Union, because its scope is Cosmetics Regulation (EC) No 1223/2009. But the practical reach is wider: INCI nomenclature has become the de facto reference vocabulary for cosmetic labelling worldwide, so a record built for EU compliance gets read by formulators on every continent. Treat it as EU law with global gravity.

Temporal - content traces to the 1976 Cosmetics Directive and carries forward through every amendment since, with entries flipping between Active and Inactive status as the annexes are revised. The corpus is therefore not a snapshot but an accretion: today's Annex III restrictions sit beside decades of accumulated inventory entries, and status is the field that tells you which obligations are current.

Granularity - one record per ingredient or regulated substance, cross-referenced to annex number, function list, restriction wording and concentration limits. There is no facility, brand or product dimension here - this is the ingredient layer, and pairing it with product-level sources is how the picture completes.

API, files, or your warehouse. Daily, weekly, or hourly.

You pick the channel and the cadence; resolution, cleaning and schema stability are our problem. Each record ships with its fields populated - INCI and US names, CAS and EC numbers, item type, annex placement, functions, chemical description, restriction wording, concentration ceilings and status - normalised across the identification, classification and compliance layers, including the sparse corners where a real record legitimately leaves a field empty. Bulk pulls land as files, continuous consumption runs through the API, and warehouse-native loads write straight into your own storage. A sample goes out first, cut to the ingredient families and annex slices you actually need.

• EU label-compliance screening - check every INCI name on a formula against Active annex entries before artwork, so a prohibited or newly restricted substance surfaces while the fix is still a reformulation rather than a recall.
• Restricted-substance watchlists - pull the Annex III slice with maximum concentration limits and alert when a supplier spec crosses a threshold the Regulation sets.
• Supplier ingredient reconciliation - join trade names to CAS and EC numbers, resolve them to canonical INCI names, and end the per-supplier naming dialects that slow formula audits.
• Formulation function mapping - the controlled function list turns ingredient searches into formulation questions: which named humectants, absorbents and viscosity controllers exist.
• Regulatory-change impact analysis - Active/Inactive status flips mark where the Regulation moved, and diffing the annex layer over time shows which product lines a revision touches.
• Claim substantiation and due diligence - citing the Commission's own record for identity, function and restriction survives legal review better than citing a retailer page.

Regulatory affairs and compliance managers screen formulations against annex restrictions and concentration ceilings without hand-copying annex documents into spreadsheets; see regulatory use cases for market researchers. R&D formulation chemists search by function and chemistry to shortlist ingredients whose INCI names will survive labelling review. Data scientists and ML engineers use CAS-to-INCI resolution and the function taxonomy as the backbone for ingredient embeddings and product-ingredient graphs; see data scientists use cases. Competitive intelligence teams track which restricted substances rivals' categories depend on; see competitive intel product teams use cases. Developers building beauty-tech products get a stable ingredient spine to hang product catalogues on; see developers builders use cases. If you need the vocabulary itself defined, start with the INCI nomenclature glossary entry.

Four things, stated plainly.

First, inventory presence is not permission. The Commission says it directly: only Regulation 1223/2009 content has legal value, cross-references between substances and ingredients are not binding, and the inventory is not a positive list. A name lookup answers 'what is this called', never 'may I use this' - the annex entry answers the second question.

Second, two record types share one schema. Inventory ingredients and regulated substances differ exactly where it matters - annex number, restriction wording, concentration limits. Filter on item type before you build anything, or your compliance logic will meet naming records and mistake them for clearances.

Third, sparse records are real records. Verified entries like PEG-20 COCAMIDE carry a CAS number but no recorded functions and no annex attachment. Blank fields are signal about how the Commission classifies an ingredient, not missing data - handle them explicitly rather than dropping rows.

Fourth, the score reflects the job. Datadory rates this record 9 out of 10 against a catalog mean of 7.81 - verified definitions across all three field layers, with the caveats above documented rather than hidden. When your question moves from ingredients to products on shelves, pair this with INCIDecoder or Open Beauty Facts. The sample shows the corpus unchanged, caveats included.

Of the 1,744 datasets Datadory catalogs across 159 viable industries, seven cover personal care products, and CosIng is the regulatory backbone among them. The companions:

• INCIDecoder / INKEEDecoder - ingredient encyclopedia joined to real product INCI lists, showing how CosIng names land on shelves.
• COSDNA Cosmetics Ingredient Analysis - community-built ingredient analysis with acuity ratings per ingredient and product.
• Open Beauty Facts - global crowdsourced cosmetics product database with full ingredient parsing.
• Sephora structured product catalog - product, rating and review structure for a major beauty retailer.
• Ulta Beauty structured product catalog - the mass-prestige counterpart: product listings with pricing and review signals.
• Cosmetics & beauty collections on Hugging Face - community ML-ready collections spanning reviews, ingredients and product imagery.

Browse the personal care products data hub for the full slice, or see the best personal-care-products datasets ranking.