On Sunday and Monday I posted about this outbreak. One post showed that FDA already answered the argument Taylor Farms is making, in a 2020 warning letter to another salad company. The other walked through every public version of the theory that the parasite came out of Michigan and Ohio water, and the papers offered to support it.

Monday, the International Fresh Produce Association posted the final version of its consumer FAQ — stamped September 14 at 10:00 a.m., with a note that the page will not be updated again and that readers should go to the federal sites. It is now the association’s permanent record of the 2026 season. Some of it deserves credit. One part of it should still be withdrawn. And on the central question my first post was about, the association and one of its largest members are now on opposite sides, with the association plainly in the right.

On the question Taylor Farms is fighting, IFPA answers it straight

Asked whether a specific food was identified, the page says yes. Asked whether the recalled product tested positive, it says there were no confirmed positive product results at any point in the investigation, that the recall and the findings rested on epidemiologic evidence and traceback rather than a positive laboratory test, that produce testing for this parasite has real and documented limitations including low sensitivity at low contamination levels, that implicated product is usually gone before testing is possible — and that the absence of a positive sample did not mean the epidemiological and traceback findings were wrong.

Compare that with the sentence still posted on the Taylor Farms hub: epidemiology can point to a potential source but cannot confirm that a specific product caused illness, and that confirmation requires a positive lab test. One of those two statements matches what FDA told Fresh Express in October 2020 — that it is not uncommon to see many negatives even when contamination has occurred. It is not the one on the company’s website.

The narrowing on water is real, and it belongs in print

In the version stamped September 4, the one I quoted on Sunday, the sewage-overflow material sat under a heading asking what else could be a source of infection. In the final version it has moved into a section describing what IFPA raised during the investigation, written in the past tense, and it is fenced: the page says those concerns were raised in connection with the smaller, still-unconfirmed clusters, not as an alternative explanation for the confirmed Taylor Farms de Mexico finding, which FDA tied to imported product from central Mexico.

That is a narrowing and I said in my last post that it belonged in print. It still does. The association has conceded the recall cluster and scoped water to the cases nobody can explain — roughly six thousand seven hundred laboratory-confirmed illnesses this season, on the agencies’ own figures, sitting outside the outbreak count with no source publicly identified. Those cases are real. Adding water questions to a hypothesis-generating questionnaire costs almost nothing and might answer something.

One claim survived the edit and should not have

The page still says there was documented, significant rainfall and combined sewer overflow activity in parts of the affected region in mid-June, and that the timing is worth continued attention. The page’s own biology answers it. IFPA states that oocysts need seven to fifteen days in a warm, moist environment before they can infect anyone, and that symptoms begin on average about a week after exposure. CDC puts the first outbreak-linked illness onset at June 14. Work backward through a week of incubation and one to two weeks of sporulation and the contamination event lands in late May. A mid-June overflow is downstream of the illnesses it is offered to explain.

Scoping that claim to the unexplained clusters does not repair it. The timing argument fails on IFPA’s own numbers no matter which cases it is pointed at. On a page that will never be updated again, it should come out.

The word doing the quiet work is subcluster

Throughout the final version, the 12,883 illnesses are a subcluster of a broader outbreak. Neither agency uses that word. CDC’s final update calls it a multistate outbreak and states in its own voice that epidemiologic and traceback data showed the Taylor Farms de Mexico iceberg lettuce was contaminated and made people sick. FDA’s advisory says the same and keeps its investigation open. Nothing is inaccurate about the word subcluster — there were other clusters — but it shrinks the largest confirmed produce outbreak in the history of this parasite into a portion of something unexplained, and vocabulary does argument work that footnotes cannot.

Two sentences on that page cut against me, and I will say so

The page says findings in a Cyclospora investigation are strongest when supported by multiple complementary lines of evidence rather than resting on any single source in isolation. And it says patient interviews and purchase records are an important piece of the puzzle, “but not a substitute for the physical and laboratory evidence” that is harder to come by with this parasite. A defense lawyer will put that second sentence next to everything else I have quoted and argue the association agrees with Taylor Farms.

The answer is on the same page. There were multiple lines here: epidemiologic interviews across twenty-one states, a traceback that converged on a single supplier, and CDC genotyping showing the parasites from sick people are genetically related. IFPA read all of that and still answered yes when asked whether a food had been identified. Wanting more evidence is not the same as saying the evidence you have proves nothing. Taylor Farms is making the second argument. Its association is not.

One error worth fixing even on a closed page

The page tells consumers that all water used for crop irrigation must meet cleanliness standards set by FDA. Three answers later it explains, correctly, that since the 2024 update to the Produce Safety Rule pre-harvest agricultural water is no longer tested against numerical criteria at all — farms perform a documented annual assessment instead, and only post-harvest water carries a testable microbial standard. Both statements cannot be true. The first is the one a worried consumer will remember, and it is the wrong one.

There is also a mismatch on the July 18 sample. IFPA describes it as routine random sampling unrelated to the recalled product. FDA’s own July 19 update describes it as a sample of lettuce supplied by Taylor Farms de Mexico. Both can be true if the lot was never part of the recall, but the supplier’s name is in FDA’s sentence and not in IFPA’s.

What the page gets right, and what I will back

Three things on that page are more useful than anything the implicated company has published all summer.

First, the undercount, stated by the industry itself: this parasite does not appear on the standard parasite panel, a physician has to ask for it specifically, and doctors are advised to test three separate stool samples on different days before calling a negative real. Every count this season is a floor.

Second, the admission that there is no FDA guidance and no validated, Cyclospora-specific protocol telling growers how to prevent contamination. IFPA calls it a real gap rather than an oversight, and points growers to university extension material for lack of anything federal. That is a striking thing for the largest produce association in the country to have to write, and it is an argument for a rule rather than against one.

Third, the FSMA 204 answer. IFPA says it cannot quantify what the delay cost this investigation but that the underlying gap is real, that traceback here ran on interviews, purchase records and manual work, and that it continues to advocate for full implementation. Good. I will take that sentence to Congress the next time somebody proposes to push the compliance date again.

The association has also proposed an FDA-led public-private task force to do a root cause analysis now that the investigation has closed. I support that, on three conditions: the findings get published, the sampling data behind them gets released, and somebody in the room does not work for the industry being examined. FDA has said it plans a root cause analysis of its own. Those are compatible.

The ask: nine things IFPA could put its name on before next May

On Saturday I listed ten things the FDA should do before the next season. This is the industry half of that list. None of it requires anyone to concede a lawsuit, and most of it does not require a statute. Several items IFPA has already half-endorsed, which is the point — finish the sentence.

1. Oppose any further delay of FSMA 204 and say so to Congress. The compliance date is July 20, 2028, moved there from January 2026, and last November’s appropriations law directed FDA not to enforce before that date. IFPA says it cannot quantify what the delay cost this investigation but that the gap is real. Then defend the date publicly and ask Congress to lift the spending prohibition that keeps FDA from building the receiving system Section 204 ordered in 2011.

2. Co-fund a detection method a grower can actually run. FDA has a validated laboratory method for finished produce. There is nothing comparable for irrigation water, sediment or soil at commercial scale and speed, and IFPA’s own page concedes there is no validated prevention protocol for this parasite at all. Put industry money behind method development, insist the limit of detection be published, and stop asking members to manage a hazard they cannot measure.

3. Back restoration of the defunded federal Cyclospora research. Two federal research projects went unfunded in the FY2026 agriculture bill and a third is being moved away from the scientists running it. An association that wants better tools should be saying which questions its members cannot answer without that work.

4. Support making Cyclospora mandatory in FoodNet again. Reporting for this parasite became optional on July 1, 2025, after twenty-eight years. We went into the worst season on record without the yardstick for normal. Restoring it costs the industry nothing and gives every member a defensible baseline.

5. Fund state laboratory capacity for genotyping. CDC’s genotyping was the first tool this summer that did not depend on somebody remembering a June salad, and it is the tool most likely to separate one company’s product from everyone else’s in the next outbreak. Industry has more to gain from that separation than anybody.

6. Adopt a members’ disclosure standard for testing claims. Any member that publicly cites its own testing numbers publishes the composition behind them: matrix, method, sample weight, dates, lots and results. FDA’s 2013 assessment of Taylor Farms de Mexico did exactly that — it printed the table, and the table showed that of 793 samples, 42 were finished product. An association can write that standard into its own code of practice tomorrow.

7. Write a worker sanitation standard built for a human-only parasite. That same 2013 assessment found field handwashing stations whose valve handles created a cross-contamination route and recommended that employees with diarrhea be encouraged to submit stool samples. Thirteen years later there is no standard requiring hands-free handwashing at field sanitary units, documented pump-out siting and service records, or access to testing and paid sick leave for symptomatic harvest workers. Humans are the only host. Everything else is downstream of that.

8. Make a documented growing-region assessment a condition of supply for imported leafy greens. For regions with a documented history of this parasite, ask members to require — under their existing foreign supplier verification obligations — a written assessment of the region’s sanitation infrastructure and water sources, and publish the audit standard so buyers and growers are working from the same document.

9. Fund wastewater surveillance for this parasite, including in growing regions. This one is IFPA’s idea and I am adopting it. The association asked CDC to consider routine wastewater surveillance to understand where the parasite is present. It is a sound proposal for an organism that exists only in human waste, and it would be more useful still in the valleys that grow American winter produce than in Lake Erie.

Add the condition I put on the task force: FDA publishes the root cause findings and the sampling results either way, and someone in the room does not work for the industry being examined. A published negative tells every farm in that valley the contamination was not where investigators looked. FDA’s letter to industry asks members to apply rules already on the books and promises to fold lessons into an action plan first published in 2021. Nine specific commitments from the largest produce association in the country would be worth more than that letter.

The association is carrying weight it should put down

IFPA’s page and Taylor Farms’ page now sit on opposite sides of the same question. One says the finding rested on epidemiology and traceback and that the absence of a positive sample did not make it wrong. The other says only a positive sample could have confirmed anything. FDA said the first thing in an enforcement letter six years ago. CDC said it in its final update this month. The association got it right.

What remains is the water paragraph, kept alive on a page that will never be edited again, on timing its own biology contradicts. Take it down and take up the nine. Nineteen thousand five hundred and ninety-five Americans were laboratory-confirmed with this parasite between May and September, and roughly six thousand seven hundred of them still have nothing written next to their names. The season starts again in May. There is time.

CDC closed the iceberg lettuce outbreak on September 11 with a final count of 12,883 illnesses in twenty-one states, at least 570 hospitalizations and two deaths in Michigan, with onsets running from June 14 to August 17. At the peak, before the recall, CDC says more than a thousand infections were occurring in a single day; by August the average was under two a day. Nationally, FDA reports that CDC had counted 19,595 laboratory-confirmed cyclosporiasis cases as of September 8. FDA’s investigation of Taylor Farms de Mexico is still open, and the samples FDA and Mexican officials collected at the growers and the processing plant are pending analysis.

CDC’s final update states the conclusion in its own voice: epidemiologic and traceback data showed that iceberg lettuce from Taylor Farms de Mexico was contaminated with Cyclospora and made people sick. There has never been a positive product test.

The company’s Cyclospora information hub, last modified September 1, answers that. Epidemiologic evidence, it says, is a legitimate tool that can point investigators toward a potential source. Then: it cannot confirm that a specific product caused illness, and that confirmation requires a positive lab test of the product. FDA has no confirmed positive result for any Taylor Farms product as of August 27. The company’s own testing, approximately two thousand samples taken in central Mexico since May, has returned zero positives. The investigation continues, and we will follow the science.

I have been reading company statements in outbreaks for more than thirty years. That paragraph was not drafted by a microbiologist. It reads like a causation defense, or like a communications team looking for a sentence that is technically defensible and lands like exoneration. Here is how you can tell, and here is why it does not survive contact with FDA’s own files.

FDA answered this exact argument in 2020, in writing, to another company

In the 2020 Cyclospora outbreak, FDA and CDC linked bagged salad mixes made at Fresh Express’s Streamwood, Illinois plant to at least 701 illnesses in fourteen states. FDA never found the parasite in a package of salad. It found it in canal water near a farm in south Florida. It then issued a warning letter dated October 20, 2020, and that letter is the best answer anyone has written to the paragraph now sitting on the Taylor Farms website.

Three things in it matter. First, the finding: based on epidemiological and traceback evidence, CDC and FDA determined that the bagged salad mixes made at that facility were the source of the outbreak. Second, the legal consequence: FDA declared the product adulterated under section 402(a)(1) of the Food, Drug, and Cosmetic Act, citing the presence of the parasite in the salad mixes as evidenced by the outbreak itself. Not by a laboratory result. By the outbreak.

Third, and this is the sentence Taylor Farms needs to reckon with. Fresh Express had handed FDA its own negative testing results. FDA wrote back that the company had provided results showing no detection of the parasite in any of its samples and then explained why that did not matter: human pathogens like Cyclosporaare not present in high numbers or homogeneously distributed, and it is not uncommon to see many negatives even when contamination has occurred.

That is FDA, in an enforcement document, telling a salad company that its own negative test results do not mean what the company wants them to mean. Taylor Farms is offering the public the same argument, with a bigger number attached.

The sentence states a legal standard, not a scientific one

Read the verb. The claim is not that the lettuce tested clean. The claim is that nothing can confirm that a specific product caused illness without a positive laboratory test. Causation and confirmation are courtroom words. Outbreak investigators do not use them that way; they identify a vehicle from the evidence that exists — interviews, case-control studies, product tracing, environmental assessment, and laboratory work when there is something left to test.

It also sets a standard higher than any court applies. In civil litigation the question is what more likely than not caused the harm. A rule requiring laboratory confirmation of the food would exceed the burden of proof in every American courtroom.

It appears in no guideline anywhere

The Council to Improve Foodborne Outbreak Response publishes the guidelines state and local health departments use. It is the joint work of the epidemiologists, laboratory directors, environmental health specialists and regulators who do this for a living, with CDC, FDA and FSIS at the table. Chapter 6, Box 6.4, sets out when contaminated food should come off the market, and lists the bases in the alternative, separated by the word OR. The first is specific exposure information linking illness to consumption of that food, even if the pathogen has not been isolated from the food. A definitive laboratory result on a representative sample is listed second, as another way to get there. Section 6.1.1 of the same chapter says waiting for laboratory results may not be necessary before control measures are implemented.

It borrows FDA’s phrase and leaves FDA’s sentence behind

No confirmed positive sample results for product testing is FDA’s wording, from its July 19 false-positive correction. Five days later, on the same advisory page, FDA wrote that the false-positive sample does not change the basis for its ongoing outbreak investigation or the overwhelming epidemiological data supporting the recall, and that traceback and outbreak data continued to converge on shredded iceberg lettuce from Taylor Farms growers in central Mexico. Both sentences are on one FDA page. The hub uses one of them.

It is frozen on a date that has been overtaken

As of August 27, was accurate when written. It is not the current state of the record. FDA updated the advisory on September 11 to 12,883 illnesses in twenty-one states with 570 hospitalizations, and CDC closed the outbreak the same day. The company’s page has not moved since September 1. The number of people has changed; the sentence has not.

It explains the limits of epidemiology and says nothing about the limits of its own testing

Two thousand samples is offered as evidence, and it is the only quantitative claim on the page that no member of the public can check. Samples of what, taken where, when, by whom, by what method, at what sample weight, from which lots? The paragraph that carefully explains what epidemiology cannot do says nothing at all about what its own testing program can do. And then the tagline: we will follow the science arrives one sentence after the page departs from it.

The company did not apply this standard itself, twice

The hub’s own timeline says FDA and CDC brought the traceback and epidemiologic data on July 15, that the company raised scientific questions about the government’s assumptions and methodology on July 15 and 16, and that on July 16 it decided to recall. On July 17 it announced a recall broader than FDA’s traceback — all iceberg sourced and processed through its central Mexico facility — and the page says it did this voluntarily, before FDA asked. There was no positive test on any of those days, and there never has been one. Under the standard the FAQ now proposes, nothing had been confirmed, there was nothing to recall, and the shredded iceberg would have stayed on the line.

The same thing happened in 2013, at the same plant. The company suspended shipments to the United States on August 9, 2013, with no positive product test in existence, and chairman Bruce Taylor told reporters at the time that all our tests have been negative and there was no evidence of Cyclospora in the product. FDA’s environmental assessment, conducted August 12 to 19 at Doctor Mora and five ranches, reports that in the combined sampling of CDC, FDA and Taylor Farms de Mexico, approximately 835 product, water and environmental samples were analyzed, including over 269 human fecal samples collected from on-ranch sanitary facilities. All tested negative except one FDA sample, number 744382, which was inconclusive because the turbidity of the recycled wash water inhibited analysis by the BAM method. The epidemiologic and traceback work still tied the salad mix from that plant to restaurant illnesses in Iowa and Nebraska; CDC’s surveillance summary counts 162 laboratory-confirmed cases among people who ate at two related restaurant chains.

A company that recalls without a positive test, twice, and then tells the public that only a positive test can confirm anything, is not describing how it thinks. It is describing what it would like a jury to believe.

Run the rule backward and almost nothing gets solved

In 1996, 1,465 cases of cyclosporiasis across twenty states, the District of Columbia and two Canadian provinces were associated with Guatemalan raspberries. As the National Advisory Committee on Microbiological Criteria for Foods recounts it, no positive results for the parasite were obtained from any of the environmental samples. In 2013, Texas investigators ran a case-control study on a restaurant cluster and found a matched odds ratio of 19.8 for cilantro, with traceback converging on three suppliers in Puebla. No positive product sample. In 2020, the Fresh Express outbreak above, where FDA found the parasite in a canal and never in a bag. Under the rule this page proposes, none of those vehicles was ever identified and nobody was ever asked to change anything.

A negative result is the expected result, even when the parasite is there

When Canadian federal scientists verified FDA’s BAM 19b method, leafy greens spiked with two hundred oocysts came back positive 93 percent of the time. Leafy greens spiked with ten oocysts came back positive 30 percent of the time. Seven out of ten known-contaminated samples read clean.

Here is the company’s best counterargument, and the produce industry does make it: FDA’s validated method can detect as few as five oocysts in a 25-gram sample. Read the fine print on what that means. Under FDA’s validation guidelines, the limit of detection is the seeding level that produces fractional results — roughly half of known-contaminated samples test positive and half do not. A method’s published floor is the point where it is a coin flip, not the point where it is reliable. Published modeling of FDA’s own validation data puts a number on it: a single 25-gram produce sample reliably detects contamination only above roughly 0.92 oocysts per gram. FDA’s 2013 assessment of this very plant estimates the infectious dose at ten to a hundred sporulated oocysts.

Do that arithmetic. A hundred-gram serving of lettuce carrying fifty oocysts is enough to put somebody in the hospital and sits well below the level at which one sample reliably reports anything. FDA called detection of this parasite complex when it walked back its own July 18 result. That complexity does not run in one direction.

The company’s own trade association says the opposite

The International Fresh Produce Association’s consumer Cyclospora FAQ, last updated September 14, asks whether a specific food was identified and answers yes. Asked whether the recalled product tested positive, it says there were no confirmed positive product results, that the recall and the findings rested on epidemiologic evidence and traceback rather than a positive laboratory test, that produce testing for this parasite has real documented limitations including low sensitivity at low contamination levels, and that the absence of a positive sample did not mean the epidemiological and traceback findings were wrong. That is the produce industry’s own association, writing for consumers, contradicting one of its largest members.

I have heard this argument in a courtroom. It lost.

In October 1998, eleven children were infected with E. coli O157:H7 after a taco lunch at Finley Elementary School in Benton County, Washington. I tried that case. The investigating team, headed by the state epidemiologist for communicable diseases, concluded in writing that because no other common school activity was identified, it was reasonable to conclude a meal served at the school was the likely source, and that the ground beef in the tacos was likely the vehicle. The team called it the most probable source, meaning greater than fifty percent certainty. The jury was asked whether the taco filling contained E. coli and whether it caused the children’s injuries and answered yes to both. The Court of Appeals affirmed on November 21, 2002.

That is what epidemiologic proof looks like when it is tested by cross-examination rather than posted on a website.

Two thousand samples of what?

We know exactly why this question matters, because the 2013 environmental assessment published the breakdown of the last time this company counted its own samples. Of the 793 samples Taylor Farms de Mexico collected across its plant and thirty ranches that year, 273 were sanitary samples, 238 were agricultural inputs, 164 were pre-harvest plant tissue, 47 were environmental, 29 were in-process — and 42 were finished product. A sample count is not a finding. The composition is the finding.

So, publish the composition. None of this is proprietary.

Samples of what. Finished product, incoming raw lettuce, standing crop, irrigation water, wash water, environmental swabs, sanitary facilities. Give the count for each, as the 2013 report did.

By what method. BAM 19b for produce and 19c for water, or an in-house assay? Was it verified for shredded iceberg specifically, and what limit of detection did that verification show?

At what sample size. Twenty-five grams each, or grabs composited into a single test? How many grams were tested against a lot of what weight?

When. The recalled product was distributed June 29 through July 16. How many of the two thousand samples came from lots produced in that window? If the answer is none, say so plainly.

Where. How many ranches, which ones, and how many samples came from the Doctor Mora plant as against the fields that supplied it.

Against what volume. The recall covered 236,192 cases of iceberg lettuce and salad mixes. Two thousand samples against a season of production is a sampling rate, and the company can state it.

Who ran them. The company’s laboratory or an accredited independent one? Will the reports, chain-of-custody records and raw Cq values go to FDA and to the public?

Were there presumptives. Any initial positives retested and reclassified, and any samples where inhibition was detected? FDA disclosed its own reversal within a day. An inhibited sample is not a negative one — FDA sample 744382 is the precedent, at this plant.

And the water. In 2013 this company committed to a sampling program covering products and water as a condition of resuming shipments. Thirteen years on, there is still no public result.

Publish the data or stop citing the number

I have been through every version of this company’s statement since July 17, and through four weeks of the public record since. The two thousand has been carried forward the whole time without ever acquiring a noun.

FDA already wrote the answer, to a different company, six years ago: it is not uncommon to see many negatives even when contamination has occurred. Twelve thousand eight hundred and eighty-three people are in CDC’s final count for this one outbreak. Five hundred and seventy went to a hospital. Two died. They are owed something better than a sentence built to survive a deposition.

I am headed to Norway. On September 23 and 24 I am giving two talks at the Aquatiq Food Forum at the Clarion Hotel Oslo, to a room of food industry leaders, Mattilsynet, Nestlé, McDonald’s Nordic, Mowi, Coop, NorgesGruppen and Norsk Kylling.

Day one is called “Why It Is a Bad Idea to Poison Your Customers.” Day two is “The Criminalization of Food Safety Failures.” They are one argument in two parts, and the second half is the part that has changed most in the last two years. The program is here: Aquatiq Food Forum 2026.

The first title gets a laugh when I say it out loud. It should not. In thirty-three years of representing people hurt by contaminated food, I have never met a company that set out to poison anybody. That is the whole point. The harm comes from a supplier nobody audited, a swab nobody took, a positive nobody acted on, a shipment somebody released because the alternative was expensive that week.

Day one: the business case

I will start where I always start, with two people. Stephanie Smith was twenty years old and teaching dance at three studios when she ate a hamburger in September 2007. She was in the hospital until June of the following year. E. coli O157:H7, hemolytic uremic syndrome, a stroke, brain injury, dialysis, transplants ahead of her, and no children. The jury numbers in her case were $1.9 million in past medical bills, $29.4 million in future care and $1.2 million in lost earnings. Notice the shape of that. The past medical bills are the smallest number. The expensive part of a catastrophic injury is the fifty years that follow it.

Linda Rivera was a mother of six who worked with disabled children. She ate raw cookie dough in May 2009 and was hospitalized until May 2011. Large intestine removed, stroke, liver failure, congestive heart failure. She missed three of her children’s graduations. She survived the outbreak and died later of its consequences.

Norway already knows this story

I do not need to import all of my examples. In the winter of 2006, Norwegian pediatricians reported a cluster of HUS cases to Folkehelseinstituttet. Seventeen people were identified with E. coli O103:H25 between February 20 and April 6. Ten children developed HUS. One of them died. The source was a traditional cured mutton sausage, morrpølse, from a single plant — 3.5 tonnes of it, sold nationwide. It took 931 food and environmental samples to find, and the strain was followed from the finished sausage back to a mutton supplier, a slaughterhouse and four farms (Schimmer et al., BMC Infectious Diseases 2008;8:41).

That is Norway’s Jack in the Box. It is worth adding the uncomfortable part, which the investigators published themselves: the first case-control study pointed at minced beef from a different producer, and the public was warned about the wrong product in February. The sausage was not identified until March 20. Their own stated lesson was to keep chasing alternative sources after you have already warned the public.

What actually destroys companies

Thirty-three years produces some pattern recognition. It is not the pathogen. Every processor will have a positive; that is not the story. It is the delay between knowing and telling, which juries and regulators count in hours. It is the denial, because attacking the health department is the most expensive sentence a company can say. It is the pattern, because one outbreak is bad luck and a file of prior noncompliances is a case. And it is the paper, because the email written to reassure a colleague becomes exhibit 14.

Most boards model one line of the cost — the verdict. Underneath it sit the recall, the plant suspension and requalification, the lost listings, and the covenants and valuation. Those four are, in practice, uninsurable, which is exactly backwards from how companies budget for this.

And then the good news

I end day one with the best thing that happened in food safety in my lifetime. Jack in the Box in 1993 produced more than 600 confirmed infections and killed four children. In 1994 USDA declared E. coli O157:H7 an adulterant in raw ground beef. HACCP followed, then testing with consequences. The 2002 ConAgra recall was the last of the great hamburger cases I litigated. Today my firm handles essentially none of them.

Government action took a large part of my practice away, and I am grateful. The industry fought that rule, called it a travesty, and sued the USDA over it. Those same companies are still in business thirty years later selling a safer product. Nothing about beef made it uniquely fixable. What has been missing for leafy greens, ready-to-eat meats and powdered infant formula is not technology. It is a clear standard and a price for missing it.

Day two: when it stops being about money

The second talk is colder. Civil cases mean a company pays and insurance absorbs part of it. Criminal cases mean a named person is charged, and insurance does not cover a prison sentence. The trigger is almost never the pathogen. It is the concealment.

Stewart Parnell of Peanut Corporation of America is serving twenty-eight years. Nine people died and 714 were infected across 46 states, but he was not convicted of killing anyone. He was convicted of what he told customers about laboratory results — the certificates, the positive retested until it came back negative, and fifteen years of ordinary internal email read aloud to a jury. Look at who else was indicted: the owner, his brother, a plant operator, a plant manager, and the quality assurance manager. That last one is the person sitting in the room in Oslo.

Intent is not required. Under the Park doctrine a person can be convicted under the Food, Drug, and Cosmetic Act without proof of knowing or willful conduct, on the basis that they held a position that would have let them prevent the violation. The Jensen brothers pleaded to a five-count information after their cantaloupes killed thirty-three people — I represented all thirty-three families — and nobody alleged they lied about anything. The DeCosters went to jail for three months each on a record with no proof of knowledge, and it was upheld on appeal.

The front line has moved to Europe

The most active food safety prosecutions in the world right now are not American. A British restaurant owner is serving six years for gross negligence manslaughter after substituting a cheaper nut mix and killing a customer who had written “no nuts” on the order. In France, Nestlé France and its subsidiary were placed under formal investigation in July 2024 over the Buitoni pizzas that gave 48 children HUS and killed two. Lactalis and its Craon subsidiary were placed under formal investigation in 2023 — and one of the allegations is failure to execute withdrawal and recall measures. Not contaminating the food. Failing to recall it properly.

That charge has no American equivalent, and it flows from architecture Norway already shares. Article 19 of Regulation 178/2002 puts primary responsibility on the food business operator and requires immediate withdrawal and notification. Norway took that regulation into the EEA Agreement in 2007, and Mattilsynet enforces it under the Food Act. A duty framed as immediate is a duty a prosecutor can measure in hours. Measure your own interval, because somebody else will.

Meanwhile America is stepping back

This is the part that will surprise a European audience. Executive Order 14294, signed in May 2025, states that criminal enforcement of strict liability regulatory offenses is generally disfavored and directs prosecution as a last resort. It does not repeal the Park doctrine. It changes who is willing to use it. The Justice Department moved to disband the Consumer Protection Branch that historically handled these cases. The Abbott infant formula investigation was closed in 2026.

And Boar’s Head, where ten people died and two members of Congress asked the Attorney General to consider charges, has produced no public answer at all. I filed a FOIA request in June asking what happened. Seventy-eight pages came back on July 2. Nothing in them says whether anyone is still working the case. What I did notice is that the law enforcement exemption USDA claimed in 2024 is no longer being cited. I cannot prove what that means, and I will say so from the platform.

The through line

Nobody sets out to poison anyone. Every person I have watched be charged believed they were managing a business problem. The evidence came from their own files, written by good people under commercial pressure. Which makes day two the same argument as day one: it is cheaper to be safe.

If you are in Oslo on the 23rd and 24th, come find me.