GHS Classification of Chemicals: How Hazard Classes and Categories Get Assigned to a Mixture

A GHS label is an output. Something upstream decided that this chemical is Acute Toxicity Category 3 rather than Category 4, and the pictogram, signal word and hazard statement followed mechanically from that decision. This page is about the decision, not the label: how a hazard class and a category actually get assigned to a mixture, using the cut-off values, the acute-toxicity additivity formula and the bridging principles that OSHA codifies in the appendices to 29 CFR 1910.1200.

If you want the label side — the nine pictograms, the difference between “Danger” and “Warning”, what H314 means — that is covered in Understanding GHS labels: pictograms, signal words and hazard statements, and the document that carries the classification to you is covered in How to read a safety data sheet. If you want the inventory of hazard classes themselves, see Common lab chemical hazard classes explained. This page assumes all of that and works one classification end to end.

First, the scoping question: does your lab classify anything?

Usually not. Under 29 CFR 1910.1200(d)(1), the duty to classify sits on the manufacturer or importer, and the standard says so in terms that are worth reading literally:

“Chemical manufacturers and importers shall evaluate chemicals produced in their workplaces or imported by them to classify the chemicals in accordance with this section. Employers are not required to classify chemicals unless they choose not to rely on the classification performed by the chemical manufacturer or importer for the chemical to satisfy this paragraph (d)(1).” — 29 CFR 1910.1200(d)(1)

A research laboratory buying reagents is a downstream user. It receives a classification, it does not make one. Its HazCom duties are the narrow set at 1910.1200(b)(3) — keep incoming labels intact, keep the SDSs accessible, train — described in Hazard Communication Standard (29 CFR 1910.1200) in the research laboratory.

There are three situations where a lab does become a classifier, and they are worth naming because they are the only ones that put this page’s arithmetic on your desk:

• You ship a chemical you produced. 1910.1200(b)(3)(iv) says a laboratory employer that ships hazardous chemicals is a chemical manufacturer or distributor for that purpose, and owes shipped-container labels under paragraph (f) and SDSs under (g)(6) and (g)(7). 29 CFR 1910.1450(h)(2)(iii) says the same thing from the Laboratory Standard’s side.
• You synthesised a new substance with no supplier classification. Nobody upstream has done the evaluation, because nobody upstream has the compound. If it leaves the lab, you are the classifier of first instance.
• You formulate a mixture for distribution. A buffer, stain, fixative or reference standard blended in-house and sent to a collaborating institution or a core facility is a mixture whose ingredients are classified but whose combination is not.

The third case is the common one, and it is the one worked below, because a mixture is where GHS classification stops being a lookup and starts being a calculation.

Which GHS, and whose — the version question that changes the answer

“GHS” is not one document with one set of categories, and conflating the UN system with OSHA’s adoption of it will produce a wrong classification.

• The UN GHS — the “Purple Book” — is published by UNECE and revised on a two-year cycle. The current edition is the eleventh revised edition (GHS Rev. 11, ST/SG/AC.10/30/Rev.11), released in September 2025, which added classification criteria for substances and mixtures hazardous because of their contribution to global warming, revised the aerosols and chemicals-under-pressure provisions, and added guidance on non-animal test methods for skin sensitisation and on simple asphyxiants.
• OSHA’s HazCom Standard is aligned to a specific, older revision. 29 CFR 1910.1200(a)(1) states that the section’s requirements “are intended to be consistent with the provisions of the United Nations Globally Harmonized System of Classification and Labeling of Chemicals (GHS), primarily Revision 7.” That alignment came from the HazCom final rule at 89 FR 44144 (20 May 2024); the section has been amended several times since, most recently at 91 FR 6760 (13 February 2026). The text described here is the text as codified and read on 26 August 2026 — check the current eCFR before relying on a compliance date.

The gap between them is not only a lag. The GHS is a set of building blocks, and OSHA has not adopted all of them. Two consequences are visible in the codified text itself:

• Some categories simply do not exist under HazCom. The standard’s own definition of a hazard not otherwise classified (HNOC) gives the example directly: an effect that “is under a GHS hazard category that has not been adopted by OSHA (e.g., acute toxicity Category 5)”. If your data land you in acute toxicity Category 5, there is no Category 5 to land in — the effect becomes an HNOC. An HNOC must be disclosed on the safety data sheet, but 1910.1200(f)(1) is explicit that hazards not otherwise classified “do not have to be addressed on the container.”
• Environmental hazards are not part of the HazCom classification at all. The words “aquatic” and “environmental hazard” do not appear anywhere in 1910.1200 or its appendices. The environment pictogram exists in the UN GHS and appears on supplier labels; OSHA does not require it. This is the same asymmetry NFPA 704 vs GHS labels describes from a different direction — two systems, overlapping vocabulary, different scope.

So the first line of any classification record should say which system and which revision you classified under. “Classified per GHS” is not a statement; “classified per 29 CFR 1910.1200 Appendices A and B, text current as of [date]” is.

The order of operations for a mixture

Appendix A.0.4.1 sets a strict sequence, and it is a sequence, not a menu. You may only move to the next step when the previous one has no data to give you:

1. Test data on the complete mixture. “Where test data are available for the complete mixture, the classification of the mixture will always be based on those data.” Always. A calculation never overrides a test on the actual product.
2. Bridging principles. Where the mixture itself is untested but there are sufficient data on the ingredients and on similar tested mixtures, the six bridging principles at A.0.5 apply.
3. Estimation from ingredients. Only if neither of the above is available do you fall back on the per-class methods — the additivity formula for acute toxicity, cut-off values and concentration limits for the rest.

There is one exception, at A.0.4.2: for carcinogenicity, germ cell mutagenicity and reproductive toxicity, mixtures are classified on information about the ingredient substances, not on the mixture as a whole, unless a case-by-case justification for the reverse can be provided. Those three classes also get a shorter bridging list — dilution, batching and substantially similar mixtures only. See Particularly hazardous substances in the lab for why those three classes are treated differently at the bench as well.

Worked example: classifying an in-house destaining solution

The mixture below is an illustrative composite, written for this page. It is not a real product and no real institution is described. Just as importantly, the ingredient classifications and toxicity values used as inputs are the ones stated for the purpose of the worked example — do not reuse these numbers for your own formulation. Step 1 of a real classification is always pulling the current values off your own suppliers’ safety data sheets, which is exactly where they will differ.

A laboratory blends a Coomassie destaining solution in 20 L batches and ships part of each batch to a collaborating lab at another institution. Because the material leaves the workplace, the lab is a chemical manufacturer for this product and owes a classification, a shipped-container label and an SDS.

Composition, weight for weight:

Ingredient	Concentration (% w/w)	Classification taken from the supplier SDS
Methanol	40.0	Acute Tox. (oral) Cat. 3; Skin Irrit. 2; Flam. Liq. 2; STOT SE 1
Acetic acid, glacial	10.0	Skin Corr. 1A; oral LD50 3,300 mg/kg (rat)
Water	50.0	Not classified

No test data exist for the blend, and no similar tested mixture is available, so step 1 and step 2 of A.0.4.1 are unavailable and the classification is built from the ingredients.

Step 1: decide which ingredients are “relevant”

A.1.3.3(a) and A.2.3.3.1 both define relevant ingredients as those present at ≥1% (w/w for solids, liquids, dusts, mists and vapours; v/v for gases). Both add the same caveat, and it is not decorative: if there is reason to suspect that an ingredient below 1% will affect the classification, it is relevant anyway. A.1.3.3(a) goes further — considering sub-1% ingredients “is particularly important when classifying untested mixtures which contain ingredients that are classified in Category 1 and Category 2.” A 0.4% impurity that is a Category 1 acute toxicant is not excused by the 1% threshold.

Here both methanol and acetic acid are above 1%. Water is relevant to the arithmetic only as a diluent.

Step 2: acute toxicity, by the additivity formula

Acute toxicity is the one health class with a genuine calculation rather than a lookup table. A.1.3.6.1 gives the additivity formula for the mixture’s acute toxicity estimate (ATE):

100 ÷ ATE_mix = Σ ( C_i ÷ ATE_i )

where C_i is the concentration of ingredient i and ATE_i is that ingredient’s acute toxicity estimate for the route being classified. (In the codified text the equation is reproduced as a graphic rather than as characters; the definitions of the terms immediately below it are the operative text.)

Three inclusion rules from A.1.3.6.1 decide what goes into the sum, and two of them are counter-intuitive:

• Include ingredients with a known acute toxicity in any category, and also ingredients “with an oral or dermal LD₅₀ greater than 2000 but less than or equal to 5000 mg/kg body weight (or the equivalent dose for inhalation).” That band sits entirely above Category 4. An ingredient too low in toxicity to be classified at all still enters the mixture calculation.
• Ignore ingredients presumed not acutely toxic — the standard’s own examples are water and sugar.
• Ignore ingredients whose only data come from a limit dose test at the upper Category 4 threshold that showed no acute toxicity.

Where an ingredient has a category but no numeric ATE, Table A.1.2 converts the category to a point estimate. For oral toxicity those conversions are 0.5 mg/kg for Category 1, 5 for Category 2, 100 for Category 3 and 500 for Category 4. Note that the point estimate is not the midpoint of the band and not the boundary — Category 3 spans >50 to 300 mg/kg and converts to 100. Using the band edge instead of the Table A.1.2 value is a common and consequential error.

For the destaining solution, oral route:

• Methanol, Acute Tox. 3, no numeric ATE on the SDS: converted point estimate = 100 mg/kg (Table A.1.2).
• Acetic acid, LD₅₀ 3,300 mg/kg: above Category 4 but inside the 2,000–5,000 band, so it is included at ATE = 3,300.
• Water: ignored under A.1.3.6.1(b).

100 ÷ ATE_mix = (40 ÷ 100) + (10 ÷ 3,300) = 0.4000 + 0.0030 = 0.4030

ATE_mix = 100 ÷ 0.4030 = 248 mg/kg

Table A.1.1 puts the oral bands at ≤5 (Cat. 1), >5 to 50 (Cat. 2), >50 to 300 (Cat. 3) and >300 to 2000 (Cat. 4). 248 mg/kg falls in the third band: Acute Toxicity (oral), Category 3.

Two things are worth noticing in that arithmetic. The acetic acid term contributes 0.0030 out of 0.4030 — three quarters of one percent of the total. Its presence changes ATE_mix from 250 to 248 and changes nothing about the category. Meanwhile the methanol term, driven by a converted point estimate rather than a measured LD₅₀, carries essentially the whole result. The classification of this mixture is a restatement of one number on one supplier’s SDS. That is a reason to record which SDS revision you used, and to re-run the calculation when a supplier updates it.

A.1.3.2 also allows classification for a single route so long as that route is followed — tested or estimated — for all ingredients and there is no relevant evidence of acute toxicity by other routes. Where there is such evidence, every appropriate route must be classified, the pictogram and signal word reflect the most severe category, and all relevant hazard statements appear.

Step 3: skin corrosion and irritation, by cut-off value

Skin corrosion/irritation does not use the acute-toxicity formula. It uses an additivity table — Table A.2.3 — on the theory (A.2.3.3.2) that each corrosive or irritant ingredient contributes in proportion to its potency and concentration:

Sum of ingredients classified as	Mixture is Skin Corrosive Cat. 1 when the sum is	Mixture is Skin Irritant Cat. 2 when the sum is
Skin Category 1	≥5%	≥1% but <5%
Skin Category 2	—	≥10%
(10 × Skin Category 1) + Skin Category 2	—	≥10%

The third row is the part people miss. A weighting factor of 10 is applied to corrosive ingredients that sit below the Category 1 trigger, so that they still contribute to an irritant classification. A worked case where it is decisive: a mixture containing 0.5% of a Skin Corr. 1 ingredient that the classifier has reason to treat as relevant despite being under 1%, plus 6% of a Skin Irrit. 2 ingredient. Row one does not fire (0.5% < 1%). Row two does not fire (6% < 10%). Row three does: (10 × 0.5) + 6 = 11, which is ≥10, so the mixture is Skin Irritation Category 2. Neither ingredient would have classified it on its own.

For the destaining solution, the sum of Skin Category 1 ingredients is 10% (the acetic acid), which is ≥5%, so the generic additivity route gives Skin Corrosion Category 1. The note under Table A.2.3 handles sub-categorisation: where the sum of 1A ingredients is itself ≥5%, the mixture is 1A; where 1A alone is <5% but 1A+1B is ≥5%, it is 1B; and so on down to 1C.

But this is precisely the case where you must not stop at the table. A.2.3.3.4 warns that particular care is needed for acids, bases, inorganic salts, aldehydes, phenols and surfactants, because many are corrosive or irritant below 1%, and it says plainly that “for mixtures containing strong acids or bases the pH should be used as classification criteria since pH will be a better indicator of corrosion than the concentration limits in Table A.2.3.” Where the additivity approach is unworkable for chemical reasons, Table A.2.4 takes over: an acid at pH ≤2 or a base at pH ≥11.5 present at ≥1% classifies the mixture as Skin Corr. 1, and other corrosive or irritant ingredients that defeat additivity classify at ≥1% and ≥3% respectively.

Acetic acid is a weak acid, so the strong-acid pH rule does not automatically displace additivity here — but the classifier has to make and record that judgment rather than let the table make it silently. This is where A.0.4.3.3 matters: in exceptional cases, conclusive data may show a hazard is not evident above the cut-off, and the mixture may be classified on that data, provided the data exclude the possibility that the ingredient behaves worse in the mixture than in the pure state and the mixture contains nothing that would affect that determination. That is a high bar deliberately — it is an evidence standard, not an argument standard.

Step 4: physical hazards do not work this way at all

Everything above lives in Appendix A, which is health hazards. Physical hazards are Appendix B, and Appendix B has no additivity formula, no cut-off tables and no bridging principles — the A.0.5 bridging principles are expressly scoped to “mixtures classified in accordance with A.1 through A.10”. Physical hazard criteria are property criteria, measured on the mixture as it exists.

For our solution the live question is flammability. B.6.1 defines a flammable liquid as one with a flash point of not more than 93 °C (199.4 °F), and B.6.3 requires the flash point to be determined by ASTM D56-05, ASTM D3278, ASTM D3828 or ASTM D93-08, or by any method specified in 29 CFR 1910.106(a)(14), or by any method specified in GHS Revision 7 Chapter 2.6. The initial boiling point is determined by ASTM D86-07a or ASTM D1078. You cannot derive the category from 40% methanol and a percentage; you measure the blend, or you use a method the standard names. Storage consequences follow directly from whichever category you land in — see flammable liquid storage cabinet requirements.

Step 5: what the classification then forces onto the label

Once the classes and categories exist, nothing further is discretionary. Appendix C allocates the label elements, and the shipped-container content is fixed by 1910.1200(f)(1): product identifier, signal word, hazard statements, pictograms, precautionary statements, and the name, address and telephone number of the responsible party. The pictogram and signal word reflect the most severe category across all the classes assigned, which in this example is driven by the skin corrosion and acute toxicity outcomes rather than by any single ingredient’s own label.

The mechanics of that translation — which pictogram belongs to which class, why the same chemical shows “Danger” and not “Warning” — are the subject of the GHS label guide. What matters here is the direction of travel: the category is decided first, on the evidence, and the label is generated from it. Working backwards from a pictogram you expected to see is how classifications go wrong.

Bridging principles, applied to the same mixture

The six bridging principles at A.0.5.1 are the step between “we tested the product” and “we calculated from ingredients”. They apply where the mixture is untested but sufficient data exist on both the ingredients and on similar tested mixtures. Applied to the destaining solution:

• Dilution (A.0.5.1.1). The lab needs a half-strength working solution: 20% methanol, 5% acetic acid, 75% water. Water has an equivalent or lower toxicity classification than the least toxic original ingredient and is not expected to affect the toxicity of the others, so the principle applies. It offers two routes: (a) classify the diluted mixture as equivalent to the original, or (b) for acute toxicity, apply the additivity formula. Take route (b): 100 ÷ ATE_mix = (20 ÷ 100) + (5 ÷ 3,300) = 0.2015, so ATE_mix = 496 mg/kg, which is in the >300 to 2000 band — Acute Toxicity (oral) Category 4. Halving the concentration moved the category, and with it the label moves from the skull-and-crossbones pictogram with signal word “Danger” to the exclamation mark with “Warning”. Route (a) would have kept it at Category 3. The two routes give different answers and the standard permits either, which is a good reason to state on the record which one you used.
• Batching (A.0.5.1.2). The next 20 L batch of the identical formulation, made by or under the control of the same manufacturer, may be assumed substantially equivalent to the tested batch — unless there is reason to believe variation has changed its toxicity, in which case a new classification is required. This is the principle that stops every production batch needing its own file.
• Concentration of mixtures (A.0.5.1.3). If a tested mixture is in Category 1 and you increase the concentration of the ingredients that are in Category 1, the resulting untested mixture is Category 1. The principle only runs upward from Category 1; it does not licence reasoning downward.
• Interpolation within one hazard category (A.0.5.1.4). Mixtures A and B, identical ingredients, both tested, both in the same category; untested mixture C has the same toxicologically active ingredients at concentrations between A’s and B’s; C is assumed to be in that same category. If A and B fall in different categories, there is nothing to interpolate.
• Substantially similar mixtures (A.0.5.1.5). For mixtures A+B and C+B where B is at essentially the same concentration in both, A in the first equals C in the second, and toxicity data for A and C are available and substantially equivalent (same hazard category, neither expected to affect B’s toxicity) — then if either mixture is classified on test data, the other takes the same category.
• Aerosols (A.0.5.1.6). An aerosol form of a mixture takes the same category as the tested non-aerosolised form, provided the added propellant does not affect the toxicity when spraying. This one is scoped to A.1, A.2, A.3, A.4, A.8 and A.9 only.

The lists are class-specific and shorter than people assume. Carcinogenicity, germ cell mutagenicity and reproductive toxicity get only dilution, batching and substantially similar mixtures. Aspiration toxicity (A.10) gets five — no aerosols — and adds a hard floor: for the dilution principle, “the concentration of aspiration toxicants shall not be less than 10%.” Check the list in the chapter for the class you are classifying rather than working from the A.0.5 master list.

Cut-off values are a floor, not the answer

A.0.4.3 is the paragraph that separates a classifier from a spreadsheet. It says outright that while the adopted cut-off values adequately identify the hazard for most mixtures, some mixtures contain hazardous ingredients below the cut-off that still pose an identifiable hazard, and in other cases the cut-off is considerably lower than the established non-hazardous level. Two obligations follow, and they run in opposite directions:

• A.0.4.3.2 — downward. If the classifier has information that an ingredient’s hazard will be evident below the specified cut-off, the mixture “shall be classified accordingly.” This is mandatory, not permissive.
• A.0.4.3.3 — upward. In exceptional cases, conclusive data may show the hazard is not evident above the cut-off, and the mixture may then be classified on that data — with the two guardrails described earlier.

A.0.4.4 adds a third judgment: the evaluator must take into account all available information about potential synergistic effects among ingredients, and may lower a classification on the basis of antagonistic effects only where sufficient data support it. The asymmetry is deliberate — suspicion is enough to classify up, evidence is required to classify down.

This is also why classification is not a risk assessment. A.0.1.1 is explicit: hazard classification considers only the intrinsic hazardous properties of chemicals, in three steps — identify the relevant data, review it to ascertain the hazards, determine whether the chemical is hazardous and to what degree. Exposure, quantity, engineering controls and use pattern are all outside it. Those belong in a risk assessment and in the chemical hygiene plan, downstream of the classification.

Ingredients of unknown acute toxicity: the 10% rule

Mixtures rarely arrive with a complete dataset, and A.1.3.6.2 has a specific, determinate procedure rather than a shrug:

• Where no ATE is available for an ingredient but a derived conversion value can be obtained — by route-to-route extrapolation with proper pharmacodynamic and pharmacokinetic data, from human exposure evidence showing toxic effects short of lethality, from other toxicity assays, or from structure-activity relationships on closely analogous substances — the ordinary formula may be used. A.1.3.6.2.2 cautions that this “requires substantial supplemental technical information, and a highly trained and experienced expert”.
• Where an ingredient of unknown acute toxicity is present at ≥1% and the mixture has not been tested as a whole, the mixture “cannot be attributed a definitive acute toxicity estimate” and is classified on the known ingredients only. A statement that x percent of the mixture consists of ingredients of unknown acute toxicity by that route is then required on both the label and the SDS.
• Where the total concentration of relevant ingredients with unknown acute toxicity is ≤10%, the ordinary A.1.3.6.1 formula must be used. Where it is >10%, the formula is corrected to adjust for the unknown fraction — the known-ingredient sum is scaled against (100 minus the unknown percentage) rather than against 100.

That 10% boundary is the single most useful number to carry out of this section. Below it, you calculate normally. Above it, the arithmetic changes and the label acquires a disclosure that tells every downstream user how much of the product was never characterised.

What a defensible classification record contains

Nothing in 1910.1200 prescribes a classification file format, but the SDS you produce under (g) has to be supportable, and the parts of the reasoning that are judgment rather than lookup are the parts that get questioned. A record that survives scrutiny states:

• Which system and revision was applied, and the date the regulatory text was read.
• The full composition with concentrations, and for each ingredient the supplier, SDS revision date and the classification or numeric value taken from it.
• Which tier of A.0.4.1 was used for each hazard class — mixture test data, a named bridging principle, or ingredient-based estimation — and why the higher tiers were unavailable.
• The arithmetic, shown, including which ingredients were included or ignored under A.1.3.6.1 and which Table A.1.2 conversions were used.
• Every judgment call: sub-1% ingredients treated as relevant, any A.0.4.3.2 or A.0.4.3.3 departure from a cut-off, the synergy/antagonism assessment under A.0.4.4, and the choice between dilution routes (a) and (b).
• Any hazard recorded as an HNOC, with the reason — typically that the effect falls under a GHS category OSHA has not adopted.

If the material stays inside your own institution, that record and the internal container label are the whole obligation; see secondary container labeling. If it ships, the record has to carry a full 16-section SDS and a compliant shipped-container label with it — the employer-side duties around that are set out in OSHA SDS requirements and, for the program that has to describe all of it in writing, the written hazard communication program.

More lab-compliance guidance sits on the lab compliance pillar.

Frequently asked questions

Does my laboratory have to classify chemicals under GHS?

Almost certainly not for purchased reagents. 29 CFR 1910.1200(d)(1) places the classification duty on manufacturers and importers and says employers are not required to classify unless they choose not to rely on the manufacturer’s or importer’s classification. A lab becomes a classifier only when it ships a chemical it produced, synthesises a substance nobody has classified, or formulates a mixture for distribution outside the lab.

What is the difference between a hazard class and a hazard category?

The class is the kind of hazard — acute toxicity, skin corrosion, flammable liquid. The category is the severity band within that class, and it is the thing the classification machinery actually outputs: the additivity formula produces a number, and the number falls into a band. Acute Toxicity is the class; Category 3 is what 248 mg/kg makes it.

What is an acute toxicity estimate (ATE)?

The value used to place a substance or mixture in an acute toxicity category. For a substance it is the LD₅₀ or LC₅₀ where available. For an ingredient inside a mixture calculation it is the LD₅₀/LC₅₀ if available, otherwise a conversion value from Table A.1.2 corresponding to a range test result or to the ingredient’s classification category. For the mixture it is the result of the A.1.3.6.1 additivity formula.

What is a cut-off value or concentration limit?

The ingredient concentration at or above which a classified ingredient triggers classification of the whole mixture for a given hazard class — for example, a sum of Skin Category 1 ingredients of ≥5% making the mixture Skin Corrosion Category 1. A.0.4.3 makes them rebuttable in both directions: an ingredient known to be hazardous below its cut-off classifies the mixture anyway, and conclusive data can show a hazard is not evident above it.

When can I use a bridging principle instead of testing?

When the mixture itself is untested but there are sufficient data on both the individual ingredients and on similar tested mixtures to adequately characterise its hazards, and the specific principle is on the list for that hazard class. Bridging sits above ingredient-based estimation and below actual test data on the complete mixture in the A.0.4.1 sequence.

Which GHS revision does OSHA’s HazCom Standard follow?

1910.1200(a)(1) states the standard is intended to be consistent with the UN GHS “primarily Revision 7”, an alignment set by the final rule at 89 FR 44144 (20 May 2024). The current UN edition is Revision 11 (September 2025), so supplier documents prepared to the UN text may reference criteria and categories that HazCom does not contain.

Do I have to classify environmental hazards under OSHA HazCom?

No. Aquatic and environmental hazards appear nowhere in 1910.1200 or its appendices. The environment pictogram is part of the UN GHS and appears on many supplier labels, but OSHA does not require its classification or its display. Waste and discharge obligations for the same chemicals come from EPA authorities instead, not from HazCom.

Sources

• 29 CFR 1910.1200 and Appendices A (health hazard criteria), B (physical hazard criteria) and C (allocation of label elements), read via eCFR for title 29 as in force 2026-08-01 and current through 91 FR 6760 (13 February 2026): ecfr.gov
• Hazard Communication final rule, 89 FR 44144 (20 May 2024).
• UN Globally Harmonized System of Classification and Labelling of Chemicals, eleventh revised edition (ST/SG/AC.10/30/Rev.11, 2025), UNECE: unece.org. UNECE’s own pages returned HTTP 403 to an automated fetch when this page was written, so the Rev. 11 publication details here are corroborated from secondary regulatory-news reporting rather than a direct read of the UNECE page — consult the Purple Book itself before relying on any specific Rev. 11 criterion.