Aflatoxin Limits in the EU, US and India: The Numbers and the Sampling Behind Them
Compliance

Aflatoxin Limits in the EU, US and India: The Numbers and the Sampling Behind Them

July 29, 202615 min read

The short version

  • The European Union sets aflatoxin limits per product category in Regulation (EU) 2023/915, and it sets two different limits for the same commodity depending on whether it will be sorted before consumption or eaten as it is.
  • The United States works differently. FDA applies an action level of 20 parts per billion for total aflatoxins in food, and 0.5 parts per billion for aflatoxin M1 in milk.
  • Aflatoxin is not evenly spread through a lot. A few badly contaminated kernels in twenty tonnes can fail a container, which is why the sampling rules demand 100 incremental samples and a 20 kg aggregate sample.
  • The EU sum of aflatoxins is calculated on a lower bound basis, meaning results below the limit of quantification are counted as zero.
  • Testing at the load port protects you far more cheaply than testing at the destination. A rejected container at Rotterdam costs the freight, the demurrage, the destruction or return, and the relationship.

Aflatoxin is the reason food import controls exist in the form they do. It is the single most consequential contaminant in the agricultural commodity trade, it is invisible, it is not destroyed by normal cooking, and the regulatory limits are measured in parts per billion. If you import groundnuts, spices, oilseeds, dried fruit or nuts, this is the number that can turn a profitable container into a total loss.

This is a working guide to the limits in the European Union, the United States and India, how the rules differ in structure and not just in numbers, and what to put in a contract so that you find out about a problem in Gujarat rather than at your own port. Every figure here is attributed so you can check it yourself.

What aflatoxin is and where it comes from

Aflatoxins are toxic compounds produced by certain moulds, mainly Aspergillus flavus and Aspergillus parasiticus. Four matter in food: B1, B2, G1 and G2. B1 is the most potent of the group, and it is classified by the International Agency for Research on Cancer as carcinogenic to humans. A fifth, aflatoxin M1, appears in milk when a dairy animal has eaten contaminated feed, and it is regulated separately.

The moulds are common in soil in warm climates. They cause a problem in two situations. The first is in the field, when the crop is stressed by drought or damaged by insects, because a broken seed coat is an open door. The second is in storage, when moisture and warmth let the mould grow on a crop that was clean at harvest. That second route is the one under human control, and it is where most avoidable contamination happens.

The practical consequences follow directly from the biology. Drying the crop fast and to the right moisture is the primary defence, insect control is the second, and once aflatoxin has formed there is no realistic way to remove it from a lot other than by physically removing the contaminated kernels. You cannot wash it out, cook it out, or dilute your way to compliance, and attempting the last one is fraud.

Why aflatoxin is a distribution problem, not an average

This is the concept that most buyers and quite a few suppliers get wrong, and understanding it changes how you write a contract.

Aflatoxin does not spread evenly through a consignment. It concentrates in a small number of individual kernels or pods, at concentrations that can be thousands of times the regulatory limit, while everything around them is clean. A twenty tonne container can be non-compliant because of a handful of nuts. That means the result you get depends enormously on where in the lot your sample came from, and a small sample taken from one bag tells you almost nothing.

European Union law addresses this head on. Commission Implementing Regulation (EU) 2023/2782 of 14 December 2023 lays down the methods of sampling and analysis for the control of mycotoxin levels in food. It applies from 1 April 2024 and it repealed the previous rules in Regulation (EC) No 401/2006, with a narrow carve-out allowing methods validated under point 4.3 of Annex II of the old regulation to continue until 1 January 2029.

The numbers in that regulation tell you how seriously the sampling problem is taken. For groundnuts, apricot kernels, tree nuts and large particle dried spices:

Lot weightSublotsIncremental samplesAggregate sample
500 tonnes and above100 tonne sublots10020 kg
Above 125 and below 500 tonnes5 sublots10020 kg
15 to 125 tonnes25 tonne sublots10020 kg
Below 15 tonnesNot subdivided10 to 100Up to 20 kg
Sampling requirements from Regulation (EU) 2023/2782, Part II.D

Each incremental sample is about 200 g, and the sublot weight may exceed the stated figure by up to 20%. For lots of 15 tonnes and above, the 20 kg aggregate sample is mixed and divided into two equal 10 kg laboratory samples before grinding, unless the product is going for further sorting or physical treatment and the whole 20 kg can be homogenised. Dried figs get their own regime under Part II.C: incremental samples of about 300 g, 100 of them, a 30 kg aggregate sample, divided into three 10 kg laboratory samples.

Read those figures again and notice what they imply. The regulator considers a 20 kg sample built from 100 separate grabs across the lot to be the minimum credible basis for a decision. So when a supplier offers you a certificate of analysis based on a 1 kg sample scooped from three bags, the certificate is not wrong exactly, it is simply not evidence about your container. Ask how the sample was drawn, from how many bags, and whether the sampling followed the regulation. That question separates suppliers who understand the risk from suppliers who are hoping.

One more consequence. Because the sampling variance is so large, two competent laboratories testing two properly drawn samples from the same lot can genuinely produce different results. That is not a scandal, it is the mathematics of a heterogeneous contaminant. It is also why your contract needs an agreed retest procedure and a named arbitration laboratory, decided before anybody is angry.

European Union limits

The maximum levels are set in Commission Regulation (EU) 2023/915 on maximum levels for certain contaminants in food, which replaced the long-standing Regulation (EC) No 1881/2006. It is organised as a table of numbered entries by food category, and the aflatoxin entries sit in section 1.1.

Two structural features matter more than any individual number.

One: the same commodity has two limits

The regulation distinguishes between product that will still be sorted or physically treated to reduce contamination before it reaches a consumer, and product that is going to the final consumer or straight into a food as an ingredient. The first gets a higher limit, on the logic that the sorting step will remove the bad kernels. The second gets the strict limit.

EntryCategoryAflatoxin B1Sum of B1, B2, G1 and G2
1.1.4Groundnuts and other oilseeds to be subjected to sorting or other physical treatment before human consumption8.0 µg/kg15.0 µg/kg
1.1.5Groundnuts and other oilseeds and processed products thereof, placed on the market for the final consumer or for use as a food ingredient2.0 µg/kg4.0 µg/kg
Aflatoxin maximum levels for groundnuts and other oilseeds under Regulation (EU) 2023/915

The higher limit is conditional, and the condition is documentary. Article 6 of the regulation requires that consignments relying on a to-be-sorted limit are clearly labelled with their intended use and carry the prescribed wording stating that the product must be subjected to sorting or other physical treatment to reduce aflatoxin contamination before human consumption or use as a food ingredient. There is also a specific provision covering oilseeds destined for crushing into refined vegetable oil, which reflects the fact that refining removes aflatoxin from the oil.

The practical trap is obvious once stated. If your consignment is labelled as to-be-sorted, you must actually sort it, and you must be able to show that you did. If you buy on the higher limit and then sell the product on unsorted, you own the non-compliance. Conversely, if you have no sorting capability, do not let a supplier quote you against entry 1.1.4 numbers. Buy against 1.1.5.

Two: the sum is calculated on a lower bound basis

The regulation specifies that maximum levels for the sum of aflatoxins B1, B2, G1 and G2 refer to lower bound concentrations, which are calculated on the assumption that all values below the limit of quantification are zero. In plain terms: if B2 and G2 come back as "not quantified", they contribute nothing to the sum rather than contributing their detection threshold.

This matters when a result sits close to the line, because a laboratory with a higher limit of quantification will produce a lower reported sum than a more sensitive laboratory analysing the same material. It is one more reason to name the laboratory and the method in the contract rather than accepting whichever report arrives.

Spices

Spices get their own entry, and the entry names specific spices rather than covering all of them. The named group, which includes Capsicum species such as chillies and paprika, Piper species such as black and white pepper, nutmeg, ginger and turmeric, carries a maximum of 5.0 µg/kg for aflatoxin B1 and 10.0 µg/kg for the sum. Dried ginger sits at the same levels. Mixtures of spices are addressed separately.

If your product is a spice that is not in the named list, it does not follow that no limit applies, and it certainly does not follow that no scrutiny applies. Cumin, for example, is not in the named aflatoxin group but is covered by the ochratoxin A entry. We work through what that means for a cumin contract in cumin seed export grades explained.

Ochratoxin A, the other mycotoxin on the same certificate

Ochratoxin A is produced by different moulds, mainly Aspergillus and Penicillium species, and it turns up on the same test panel. Under Regulation (EU) 2023/915, dried spices carry a maximum of 15 µg/kg for ochratoxin A under entry 1.2.17, with Capsicum species set at 20 µg/kg. It is a storage toxin more than a field toxin, so it is a fairly direct report on how the material was warehoused.

The full contaminants table also covers tree nuts, dried fruit, cereals, maize, rice, milk, and foods for infants and young children, each with its own entries and its own numbers, several of them stricter than the ones above. If you are buying outside the categories in this article, read the entry that applies to your product rather than assuming a neighbouring one is close enough.

United States limits

The United States takes a structurally different approach. Rather than legislated maximum levels per food category, the Food and Drug Administration applies action levels, which are the concentrations at which the agency will consider a product adulterated and take enforcement action.

  • 20 parts per billion for total aflatoxins in food intended for humans. Parts per billion is the same as µg/kg, so this sits above the EU's strict category limits and below the EU's to-be-sorted limits in the case of the sum.
  • 0.5 parts per billion for aflatoxin M1 in milk, which is far stricter and reflects that milk is consumed by children in quantity.
  • Higher action levels for animal feed, which vary by species and by the age and purpose of the animal, on the basis that a mature beef animal tolerates what a dairy cow or a young animal does not.

The enforcement mechanism differs too, and it is the part exporters underestimate. FDA can place a product, a firm or an origin under an import alert, which allows detention without physical examination. Once you are on it, every consignment is held by default and you carry the burden of proving compliance with private laboratory analysis, consignment by consignment, for as long as it takes. Getting removed requires a documented history of clean entries and a petition. The financial damage is rarely the one rejected container, it is the year of detentions afterwards.

If you are importing into the United States, note also that the Foreign Supplier Verification Program puts the verification obligation on you as the importer, not on your supplier. You are expected to have evaluated the hazard, which for groundnuts and spices means aflatoxin, and to have documented the controls you rely on. A supplier certificate you never assessed does not discharge that duty.

India limits

Domestically, aflatoxin in India is regulated by the Food Safety and Standards Authority of India through its contaminants, toxins and residues regulations. The general limit commonly cited for total aflatoxins across most food categories is 15 µg/kg, with a separate and much lower limit for aflatoxin M1 in milk. Because that instrument is amended regularly, treat any figure you read online, including this one, as a pointer rather than an authority, and check the current consolidated regulation before you rely on it in a contract.

For export consignments the more relevant point is that Indian export testing sits with a defined set of bodies. Depending on the commodity, that means the Spices Board for spices, APEDA for a range of agricultural products, and the Export Inspection Council and its laboratories for mandated pre-shipment inspection into certain markets. Ask which route applies to your commodity and destination, and ask to see the certificate rather than being told it exists.

For markets outside the EU, the US and India, the usual reference point is the Codex General Standard for Contaminants and Toxins in Food and Feed, CXS 193-1995, which many national regimes align with in whole or in part. Your destination may also have its own stricter national standard, so confirm rather than assume Codex applies.

How the limits compare, and why the comparison is misleading

Put side by side, the EU's 4.0 µg/kg sum for groundnuts going to a consumer looks dramatically stricter than the FDA's 20 ppb total. That comparison is technically correct and practically incomplete, because the two systems differ in what they measure, how they sample, and how they enforce. The EU regulates B1 separately and the sum separately, applies a formal sampling regulation, and operates a border control regime with defined check frequencies. The FDA operates action levels with broad discretion and a detention mechanism that bites hard once triggered.

The operational conclusion is not "the EU is harder". It is that you need to know which regime your consignment will meet before you fix the specification, because a lot that is comfortably compliant for one destination can be non-compliant for another, and you cannot redirect a container after it has been tested and found high. Decide the destination first, then buy to it.

What to actually do about it as a buyer

Test before it sails, not after it lands

This is the whole game. A pre-shipment test on a properly drawn sample costs a few hundred dollars. A rejected container at destination costs the ocean freight, the demurrage while the dispute runs, the destruction or return freight, possibly a border control listing that raises the check frequency on every subsequent consignment from that origin, and the commercial relationship. The asymmetry is enormous and it points in one direction.

Specify the sampling, not just the limit

A limit without a sampling protocol is unenforceable. Write into the contract that sampling follows Regulation (EU) 2023/2782 where the destination is the EU, or an equivalent named protocol otherwise, and state the number of incremental samples and the aggregate sample weight. Require the sampling to be witnessed, either by a third party inspector or by your own representative, and require photographs.

Name the laboratory and the method

Require ISO/IEC 17025 accreditation for the specific test, not just for the laboratory in general. Ask which method was used. High performance liquid chromatography with fluorescence detection after immunoaffinity clean-up, or liquid chromatography with tandem mass spectrometry, are the confirmatory methods. Rapid ELISA kits have a legitimate role as a screening tool on the processing line, and no place as the basis for a certificate of analysis on a shipment.

Agree how measurement uncertainty is handled

A result of 15.2 µg/kg against a limit of 15.0 is not a straightforward failure. Analytical results carry an expanded measurement uncertainty, and the EU sampling and analysis rules require that uncertainty and recovery correction be taken into account when deciding whether a lot conforms. Write into your contract how borderline results are handled, whose laboratory arbitrates, and who pays for the retest. Settle it in advance, because there is no fair way to settle it once one party has a financial stake in a specific answer.

Look at the supply chain, not just the certificate

A certificate tells you about a sample. What tells you about the next twenty containers is how the material is dried and stored. Worth asking:

  • How is the crop dried, and what moisture is it brought down to, and how fast after harvest.
  • Is there mechanical drying capacity, or does everything depend on the sun and therefore on the weather.
  • How is the warehouse managed: is it ventilated, is moisture monitored, is it pest controlled.
  • Is there colour sorting and density sorting capability, and can they show you the reject stream. A processor who cannot show you rejects is not sorting.
  • Is there any optical or fluorescence-based sorting for the commodities where it helps, such as groundnuts.
  • What happens to rejected material. The honest answer involves feed or crushing under the correct controls, not a quiet return to the food stream.

How we handle it

For the commodities in our range where aflatoxin is the live risk, principally groundnut kernels and the chilli products, we test on the shipped lot at an accredited laboratory and send you the report with the lot number on it. We will tell you which limit we have quoted against and whether it assumes sorting at your end. If you tell us the destination, we will tell you plainly whether the material we have meets it, including when the answer is no.

If you want to understand what happens at the European border once the consignment arrives, that mechanism is set out in how EU border controls work for Indian food imports. You can also send us a specification and we will price it as written rather than as we wish it were written.

Frequently asked questions

What is the aflatoxin limit in the European Union?

It depends on the product category and on whether the product will be sorted before consumption. Under Regulation (EU) 2023/915, groundnuts and other oilseeds to be sorted or physically treated before human consumption are limited to 8.0 µg/kg for aflatoxin B1 and 15.0 µg/kg for the sum of B1, B2, G1 and G2 under entry 1.1.4. The same commodities placed on the market for the final consumer or as a food ingredient are limited to 2.0 µg/kg B1 and 4.0 µg/kg for the sum under entry 1.1.5.

What is the aflatoxin limit for spices in the EU?

The named spice group, which includes Capsicum species such as chillies and paprika, Piper species, nutmeg, ginger and turmeric, is limited to 5.0 µg/kg for aflatoxin B1 and 10.0 µg/kg for the sum. Dried ginger is at the same levels and mixtures of spices are addressed separately. Dried spices also carry an ochratoxin A limit of 15 µg/kg, rising to 20 µg/kg for Capsicum species.

What is the FDA aflatoxin limit?

FDA applies action levels rather than legislated maximum levels. The action level is 20 parts per billion for total aflatoxins in human food and 0.5 parts per billion for aflatoxin M1 in milk. Animal feed action levels are higher and vary by species and animal purpose. Enforcement runs through import alerts allowing detention without physical examination, which is the part that causes lasting commercial damage.

What is the aflatoxin limit in India?

Aflatoxin is regulated by FSSAI through its contaminants, toxins and residues regulations, with a general limit for total aflatoxins commonly cited at 15 µg/kg across most food categories and a much lower limit for aflatoxin M1 in milk. Because the instrument is amended regularly, check the current consolidated regulation before relying on any figure in a contract.

How is aflatoxin sampling done for export consignments?

For groundnuts, tree nuts, apricot kernels and large particle dried spices, Regulation (EU) 2023/2782 requires 100 incremental samples of about 200 g each, combined into a 20 kg aggregate sample, with lots above 15 tonnes divided into sublots that are sampled separately. For lots of 15 tonnes and above the aggregate sample is divided into two 10 kg laboratory samples before grinding. Dried figs use 300 g increments and a 30 kg aggregate split into three laboratory samples.

Why do two laboratories give different aflatoxin results for the same lot?

Because aflatoxin concentrates in a small number of individual kernels rather than spreading evenly, so the result depends heavily on which kernels ended up in the sample. This sampling variance is inherent to the contaminant, which is why the sampling rules demand 100 grabs across the lot and why your contract needs an agreed retest procedure and a named arbitration laboratory.

Can aflatoxin be removed from a contaminated lot?

Not by washing or cooking. Aflatoxin is heat stable at normal food processing temperatures. The only practical route is physically removing the contaminated kernels through colour, density and optical sorting, which is why the EU sets a higher limit for consignments that will still be sorted. Refining removes aflatoxin from vegetable oil, which is why oilseeds destined for crushing are treated separately.

What does the lower bound calculation mean for the sum of aflatoxins?

Regulation (EU) 2023/915 specifies that the sum of B1, B2, G1 and G2 is calculated on the assumption that any value below the limit of quantification counts as zero. So non-quantified congeners add nothing to the total. It matters on borderline results, because a less sensitive laboratory will report a lower sum than a more sensitive one on identical material.

Check it yourself

The regulatory figures above come from the primary texts. They are free to read, and because the annexes are amended periodically the consolidated version is the one to cite.

Read next

Share this article