Dry stops growth. It does not kill.
For everyone
Bacteria need available water to multiply, and chocolate has very little. That is why chocolate does not spoil the way cream does. It is also the whole difficulty: bacteria that arrive in a dry food do not grow, and do not die either.
The regulator's guidance draws its line at a water activity of 0.85 and names chocolate, milk powder, nut butters and processed tree nuts among the foods below it. Salmonella has been measured surviving in chocolate for a year.
“While LMRTE foods generally do not support the growth of environmental pathogens, environmental pathogens can survive in many LMRTE foods and cause foodborne illness.” (source 1)
Measured in: Sun et al. 2023; Podolak et al. 2010
This is not: That chocolate is a risky food. Outbreaks are rare events against an enormous volume eaten; the point is why they are possible at all.
The danger is what happens after the heating step
A manufacturer’s concern
Roasting can remove a great deal of Salmonella from cocoa beans, and has been measured doing so. Nothing later in chocolate making does the same: conching is only weakly lethal, and dry chocolate protects bacteria from heat.
So anything that reaches the product after the roaster stays in it. That includes ingredients that are never heated again in the chocolate factory, such as milk powder, nuts and dried fruit, and anything picked up from the plant itself.
“The contamination of foods with Salmonella during manufacturing has occurred through transfer of Salmonella that has become established in the production environment of the plant and through the addition of contaminated raw materials or ingredients” (source 1)
Measured in: Yan et al. 2021; Nascimento et al. 2012; Zhao et al. 2024; de Lucena et al. 2026; Sekhon et al. 2021
This is not: A roasting specification. How much a given roast removes has to be established for the equipment and the product; a published figure does not transfer.
Why water is the enemy in a chocolate factory
A manufacturer’s concern
In most kitchens, cleaning means water. In a dry-food plant water is what turns a surviving contaminant into a growing one, and carries it from place to place. The guidance lists where the water comes from: cleaning itself, condensation, leaking valves, backed-up drains, roof leaks, even a fire sprinkler.
A laboratory study found that Salmonella formed films on dry steel and plastic after a wet spell, and that how long the surface had been wet was what mattered most.
“Water, even if present in very small amounts for short, sporadic time periods, can enable environmental pathogen growth and facilitate transfer of environmental pathogens from one area to another.” (source 1)
Measured in: Vaz et al. 2026
This is not: That water is never used. The guidance allows controlled wet cleaning, kept to a minimum, in a separate area where it can be, and followed by complete drying.
Dry cleaning, and what it can and cannot do
A manufacturer’s concern
Dry cleaning means removing residue without water: vacuuming, brushing, scraping, wiping. In chocolate it also means pushing the next product, or cocoa butter, through the line to carry the last one out.
Cleaning removes residue. It is a separate thing from sanitising, which kills what is left. A model of a dry line found that wiping and flushing reduced the amount of contaminated product without reliably bringing it to nothing, and pilot trials with milk chocolate show how much product a flush takes.
“At this time, we are not aware of scientific or technical evidence that demonstrates that cleaning techniques alone, including material flush techniques, are effective in mitigating pathogen contamination” (source 1)
Measured in: Daeschel et al. 2025; Zhang et al. 2024
This is not: That dry cleaning is useless. It is the right method for a dry plant; it is not, by itself, evidence that a pathogen has gone.
How contamination travels: people, wheels, air and product
A manufacturer’s concern
Once a pathogen is inside a plant it moves. The guidance lists the routes: in the food itself as it passes down the line, on hands and shoes, on the wheels of carts and forklifts, in splashed water, and in dust carried by air handling.
This is the reasoning behind keeping the raw side of a factory, where beans arrive and are cleaned, apart from the side where finished chocolate is exposed, and behind controlling who and what crosses between them. The practice is usually called hygienic zoning. The international code of practice that describes it could not be opened for this page, so the layout of zones is not described here.
“tracks the pathogen from the contamination site to other sites on shoes” (source 1)
Measured in: Podolak et al. 2010
This is not: A zoning plan. Which areas a plant separates, and how, depends on the building and the products.
Niches: where a contaminant settles in
A manufacturer’s concern
The guidance distinguishes strains that routine cleaning removes from strains that become established in equipment or in the building. The second kind lives where residue collects and cleaning does not reach: dead ends, hollow spaces, ledges, cracks and worn surfaces.
It treats finding the same strain on occasions at least 60 days apart as a sign that one may have taken up residence.
“avoid surfaces (such as dead ends, hollow spaces, flat ledges, or sharp bends/angles in product conveyance lines) where food particles, dirt, and organic matter could accumulate” (source 1)
This is not: A statement about any factory. It describes how persistence happens where it does.
Looking for it: environmental monitoring
A manufacturer’s concern
Because the product cannot be tested into safety, plants test the place instead: swabbing surfaces and surroundings on a plan, expecting to find something sooner or later, and treating a positive as information. The guidance calls the approach “seek and destroy”.
Results are read as trends. A rising share of positives, or the same area testing positive on separate occasions with clean results between, points to something not under control.
“A well-designed environmental monitoring program promotes a “seek and destroy” (also referred to as “find and fix”) approach” (source 1)
This is not: A sampling plan. Where and how often to swab is for the plant to work out and its regulator to judge.
Why a negative test on the product is weak evidence
For everyone
In the outbreaks on this page the contamination was a few cells per hundred grams, unevenly spread. A sample from a batch like that will usually test clean. In 2022 a factory found the outbreak strain in its own checks, acted, tested product negative and shipped it.
This is why “every batch is tested” is a weaker assurance than it sounds, and why regulators ask what controls are in place and not only what the tests said.
“finished product testing alone is not adequate as verification that an environmental pathogen hazard has been controlled” (source 1)
Measured in: Hockin et al. 1989; Kapperud et al. 1990; Laisnez et al. 2025
This is not: That testing is pointless. The guidance calls it most useful alongside records and environmental results.