The Mould Was Already There

Researchers at the University of Parma, working with Cranfield University, bought 212 plant-based products off UK shelves — 92 meat alternatives and 120 beverages. Burgers, vegan sausages, vegetarian chicken pieces, oat and almond and soy milks. They screened for 19 mycotoxins — fungal metabolites that ride along on grains, legumes, and seeds when mould gets into them in the field or in storage.

They found at least one in every single sample. Not most. All 212.

Then, in the same paper: every concentration sat below EU guideline levels.

Both sentences are true, they point in opposite directions, and the gap between them is where the entire story lives.

Start with the 100 percent. It is the number that will travel, and it is the number that means the least on its own. A detection frequency is a statement about the instrument as much as the food. Modern mass spectrometry finds things at parts per trillion — concentrations at which a compound is present in the way a single grain of sand is present on a beach. Run the identical products through a 1995 lab and most of them come back clean. The food would not have changed. The eyes would have.

This is the ordinary and slightly vertiginous condition of contemporary science: our ability to detect has outrun our ability to interpret. "Present" and "matters" used to be roughly the same word. They have been quietly divorcing for about thirty years and nobody sent a card.

The second thing the number can't tell you is how it compares. Nobody bought 212 beef burgers and ran them through the same machine. Without that, the study cannot say alt-meat is dirtier than the thing it replaces — and the researchers do not say so. What it can say is narrower and, I'd argue, more interesting: food made mostly of grain, legume, and seed carries the chemistry of grain, legume, and seed. Which is also true of bread. And beer. And coffee, peanut butter, maize, and the corn that becomes the feed that becomes the beef. Andrea Patriarca, one of the authors, puts it plainly: mycotoxins occur naturally in foods and cannot be completely avoided.

Now the sentence I would rather skip, because it cuts against the reassurance I have just spent four paragraphs building.

"Below EU guideline levels" does not mean one thing. For most of that 19-toxin panel it means roughly what you'd hope: the limit is set beneath a dose at which nothing measurable happens, and under it the toxicologists genuinely think you're fine. Aflatoxin B1 — named in this piece's own first sentence — does not work that way. It is a Group 1 human carcinogen, it is genotoxic, and regulators have deliberately declined to set a tolerable daily intake for it, because no threshold of safety is assumed to exist. Its EU limit is set on an ALARA basis: as low as reasonably achievable, given that you still have to have a food supply at the end of it. That is a floor on what is practical, not a line below which nothing happens.

The distinction matters here more than it usually would, because eliding it is precisely the move this piece is about to prosecute.

An ingredient label is a manifest of intentions. It records what someone decided to put in. It has no vocabulary at all for what arrived — the fungal metabolite that got on the oats in a wet August, the trace of everything the world contributed while nobody was watching. We read that list as a complete account of the object. It is a complete account of the decisions.

And the word doing the heavy lifting in this whole category is clean. Cleaner protein, clean label, clean eating. It is marketing language quietly borrowing the authority of a laboratory term, and it survives only as long as nobody points a good enough instrument at it. Now someone has, and the answer is: nothing is clean, at this resolution, ever. Not the plant version, not the animal version, not the organic version, not the one your grandmother made.

If you believed a purity was available and you had finally bought it, that is bad news. If you already suspected that eating is an act of continuous, negotiated, mostly-fine exposure to a world that has been doing chemistry for a very long time without consulting you — you were right, and dinner is still fine.

But still is doing work in that sentence I am not sure it can carry.

The concern the researchers actually raise is not any one burger. It is accumulation: repeated small exposures across a dietary pattern, over years, to compounds implicated at high doses in liver and kidney damage. That is a question about diets rather than products, which is exactly the level at which labelling does not operate. It is also, less obviously, a question about time — because the thing being accumulated is not holding still.

The mycotoxin burden is not a constant. Aspergillus and Fusarium colonisation tracks heat and humidity at flowering, at harvest, in storage, and the geography of that is moving. Paolo Battilani and colleagues modelled aflatoxin B1 in European maize under a +2 °C scenario and found a dramatic northward shift in the risk zones — contamination becoming a food safety problem in growing regions whose seasons never used to produce it.

So there are two curves running at once here, and they are the reason the study is more interesting than its headline. The detection floor is dropping, year over year, as the instruments get better. And the actual load is moving, as the climate that governs the mould changes. From inside, those produce the same graph. A rising signal and a better microscope look identical on the page.

That is the part worth holding, and it is not consoling. "Nothing was ever clean" is true, and permanent, and asks nothing of anyone — which is exactly why it is the easier ending. The harder one is that we are acquiring the ability to see the background at almost precisely the moment the background begins to move, and we will have real trouble telling the two apart.

The mould was always there. We finally built a machine sensitive enough to say hello — and not quite sensitive enough to tell us whether it's moving.