Most contaminant conversations in the spice trade are about something that was added: a pesticide, a fumigant, a processing aid. Pyrrolizidine alkaloids are different, and the difference explains why buyers keep being caught out by them. Nothing was added. The toxin arrived in the field, inside a plant that was never meant to be harvested.
A weed problem wearing a chemistry label
Pyrrolizidine alkaloids are natural toxins produced by certain plant species. They are not made by oregano, or by cumin, or by any of the herbs and spices where they turn up. They get into a batch because those plants grow among the crop and are cut, dried and milled along with it.
That origin has three consequences a buyer should internalise. It cannot be washed off, because it is inside plant material that is now part of the lot. It cannot be milled out, because milling distributes it. And it has nothing to do with how carefully the processor operates — by the time material reaches a mill, the question has already been decided in the field.
This reframes where the risk actually sits. Auditing the processing plant is not the intervention. Understanding the harvest — the field, the weed pressure, the cleaning step before drying — is.
The limits, and what the exceedances look like
Regulation (EU) 2023/915 sets maximum levels for a defined group of pyrrolizidine alkaloids across botanical goods, ranging from 1.0 to 1000 µg/kg depending on the product. For dried oregano and for mixtures consisting exclusively of dried herbs the maximum is 1000 µg/kg. For cumin as a seed spice it is 400 µg/kg.
The instructive part is not the numbers but the size of real failures. Among cases reported through the EU alert system in 2023, an organic oregano sample exceeded the permitted level by roughly sixty times, and a ground cumin sample by roughly twenty-four times.
A twenty-fold or sixty-fold exceedance is not a batch that drifted over a threshold. It is a batch containing a concentrated pocket of contaminated material. That distinction matters because it tells you what kind of control you need: not tighter process control, but detection of a rare and unevenly distributed component.
Why the sampling plan carries more weight than the analysis
If contamination were uniform, any sample would answer the question. It is not uniform. A relatively small quantity of highly contaminated plant fragments, scattered unpredictably through a large lot, can lift the lot average past the limit while the great majority of the material is clean.
That creates two symmetrical failure modes. A sample can miss the contaminated fraction entirely and certify a lot that is genuinely non-compliant. Or it can happen to catch a concentrated pocket and reject a lot whose true average is acceptable. Both are expensive, and neither is a laboratory error.
The defence is a sampling plan designed for heterogeneity — enough increments taken from enough points, properly combined and homogenised before analysis — rather than a single grab sample sent to a good laboratory. When a supplier presents a certificate of analysis for this contaminant, the sampling protocol behind it is the part worth reading. A certificate from an unrepresentative sample is a document, not evidence.
The organic assumption, stated plainly
Buyers frequently assume that an organic lot is lower risk here. The opposite is closer to the truth, and it is worth being direct about because the assumption is widespread.
Pyrrolizidine alkaloids come from weeds growing among the crop. Organic cultivation restricts the tools available for controlling those weeds, which means weed pressure — and therefore the chance of co-harvesting — can be higher rather than lower. One of the most severe reported exceedances involved an organic sample. Organic certification answers questions about cultivation inputs and chain of custody; it does not answer this one.
Nothing in this argues against organic sourcing. It argues against treating an organic certificate as a substitute for a pyrrolizidine alkaloid result.
What belongs in the brief
Three things, before production rather than after. The parameter itself, written into the specification with the destination's applicable maximum level. The sampling plan, described rather than assumed. And the analysis, with the method and the laboratory identified.
TeraVella works in spices as a sourcing partner rather than a processor — we do not claim to own a mill. What we do is verify a source against your brief, define the batch analysis parameters up front with the destination requirements in mind, and get the specification written down before a lot is produced. We hold ISO 9001, ISO 22000 and ISO 27001 and make no organic or other certification claim; where a buyer requires certified product, those documents come from the source. Where a requested timeline no longer allows for representative sampling and analysis, we say so rather than shipping and hoping. Volumes and prices are confirmed at quotation.
