This page describes how a MolGod document is put together: which datasets stand behind it, what is checked automatically, what a person does, and — the part that matters most — what happens when the evidence is not there.
The rule that governs everything below
No data means no claim. When a substance is absent from a dataset, the document does not say “no restrictions apply”. It says the status is not established and tells you to verify it. This is not a stylistic preference; it is enforced in the code that assembles the document, because a false “unrestricted” is the one error in this field that can put someone in front of a regulator with a document that appears to clear them.
The same rule governs individual values. A field with no recorded source is marked as having none rather than filled with a plausible number. Where a value is estimated rather than measured, it says so.
Where the data comes from
Regulatory layer — versioned files, not live lookups
Classification and regulatory status are read from versioned datasets held with the engine, not fetched from an API at the moment a page is rendered. This is deliberate: a document must be reproducible, and a page that depends on a third-party API being awake produces a different document on a different day.
- CLP Annex VI — harmonised classification. The harmonised entries of Regulation (EC) 1272/2008 Annex VI, including index number, hazard classes and hazard statement codes. Where a substance carries a harmonised entry, that entry is authoritative and overrides any consensus assembled from other sources.
- CLP specific concentration limits, M-factors and ATE values. The companion table to the above, used where classification depends on concentration.
- REACH Annex XVII — restrictions. Restriction entries with the substance and a summary of what is restricted.
- ADR — dangerous goods, Table A. UN number, class, packing group and proper shipping name. Entries that are explicitly “not regulated” are recorded as such, because “not listed” and “listed as unregulated” are different facts.
- Regulatory inventories. Status entries across seven lists in five jurisdictions: REACH Annex XIV and the SVHC candidate list (EU), the SVHC list (UK), TSCA (US), DSL and NDSL (Canada) and AICS (Australia).
Substance layer — public scientific databases
- PubChem (NCBI / U.S. National Library of Medicine) — identifiers, structures and physicochemical values. Public domain.
- NIST Chemistry WebBook — thermophysical values and spectra.
- Crossref — bibliographic metadata and DOI resolution for cited literature.
- Open Library (Internet Archive) — bibliographic metadata for cited books.
- Wikidata — cross-references between identifier systems.
- Health Canada Drug Product Database — where a substance also has a pharmaceutical identity.
Identifiers are resolved across sixteen systems — CAS, EC, InChI, InChIKey, SMILES, PubChem CID, ChemSpider, ChEMBL, KEGG, DrugBank, MeSH, HMDB, UNII, NSC, Wikidata and others. A card shows how many of the sixteen were resolved for that substance, so you can see at a glance how well characterised it is rather than being left to assume.
Literature
Cited literature carries a DOI wherever a DOI exists, and each cited source carries the date on which it was read. References are rendered in Chicago style so that a third party who has never contacted us can locate the source from the citation alone.
How a document is assembled
- Identity is fixed first. The CAS number is validated against its check digit. A number that fails is not silently corrected — it is marked, because a plausible-looking CAS number that does not exist is the single most damaging error a chemical document can carry.
- Classification is read against the harmonised entry. Where CLP Annex VI has an entry for the substance, it governs. Where it does not, the classification is assembled from the cited sources and labelled as such.
- Values are attached to sources. Each physicochemical value carries the source it came from, or an explicit marker that it has none.
- Gaps are recorded as gaps. Absent values are shown as absent, with pointers to where a reader can look further.
- The draft is watermarked. A document that no person has reviewed renders with a visible working-draft watermark. There is no route to an unwatermarked document that skips the review — it is an interlock in the code, not a promise in the terms.
- Review, for the reviewed product. A qualified scientific reviewer reads the document section by section against the Annex II requirements, takes a position on the content, and records the review date. That review is what removes the watermark.
What is checked — and what is not
Checked: that the identifiers are internally consistent; that the classification matches the harmonised entry where one exists; that every value shown either carries a source or is marked as unsourced; that the sixteen sections required by REACH Annex II are present; that hazard and precautionary statement codes correspond to the classification.
Not checked, and it cannot be: what is physically in your container. We do not test your material. We cannot confirm its composition, its concentration, its impurity profile or its manufacturing route. A document describes a substance; your material is a specific batch of it, and the difference between the two is exactly the part that only you can supply.
Also not checked: whether the document satisfies a particular inspector in a particular jurisdiction on a particular day. No official certification scheme for safety data sheets exists in the European Union, so no document from anyone — us included — can carry one.
Versioning
Every document carries a version number and the date on which it was compiled. The classification layer carries the version of the CLP adaptation it was read against, so that a document can be checked against the state of the law at the time it was produced. When an adaptation to technical progress changes a harmonised entry, the affected documents are identifiable by that version marker rather than by guesswork.
What this does not make you
Under REACH the supplier placing a substance or mixture on the market is responsible for its safety data sheet. A document prepared here supports that assessment. It does not transfer the obligation, it is not a legal approval, and buying it does not make a company compliant. Where that distinction matters to your situation, the honest answer is that a document is one input to your own assessment and not a substitute for it.
MolGod.org is a trading name of Nonsensia Ltd, registered in England and Wales under company number 15165427, 124–128 City Road, London EC1V 2NX, England, United Kingdom.
