251 | From the frontlines of the Cyclospora wars | Food fraud risk with co-manufacturers |

Plus: Can AI tools do food safety safely?

Aug 10, 2026

∙ Paid

This is The Rotten Apple, an inside view on food fraud and food safety for professionals, policy-makers and purveyors. Subscribe for insights, latest news and emerging trends straight to your inbox each Monday.

Subscribe to The Rotten Apple

Free food safety resources for subscribers
Cyclospora from the frontlines
Outsourced: Food fraud and co-manufacturers
AI and food safety
Food fraud news, emerging issues and recent incidents

🎧 Listen 🎧

Hi everyone!

Welcome to Issue 251. A special shoutout to 👏👏Hilde from Norway👏👏 for upgrading to a paid subscription. Your subscription helps keep this publication independent and ad-free. Thank you.

I’m delighted to present a guest post in this week’s issue, from the frontlines of the still-unfolding Cyclospora scandal in the United States. There are now two confirmed deaths connected to one of the multiple outbreaks, and many questions as yet unanswered (it was fascinating to read about what attendees at a big US food safety conference were saying in the corridors last week: Bill Marler on the Cyclospora root cause)

Also in this issue: how to reduce food fraud risks when your products are made by co-manufacturers, and the proof we (didn’t) need: AI is bad at food safety. Plus food fraud news and emerging risks for paying subscribers.

Have a great week, and thank you for being here.

Karen

Love this newsletter? Tell your friends, and help grow our global food safety community

We Didn’t Wait for the Recall Notice

A guest post by Santoshi Muriki, Food Safety Manager

I found out about the Cyclospora outbreak the same way most people probably did — a headline on TV, a story surfacing on my LinkedIn feed, and then, not long after, an actual FDA/CDC alert.

On July 16, the agencies announced a five-state outbreak linked to Taco Bell locations: 1,644 confirmed illnesses, 94 hospitalisations, across Indiana, Kentucky, Michigan, Ohio, and West Virginia. FDA’s traceback work pointed to a single supplier of Mexican-grown iceberg lettuce as the likely source. They hadn’t named the supplier yet. Taco Bell had already said it would stop sourcing them.

As Food Safety Manager at Affiliated Foods, a wholesale grocery cooperative supplying a network of more than 1,000 vendors, that unnamed part is the part that matters to me. An outbreak with no supplier’s name yet is not a reason to wait. It’s a reason to start asking questions.

An outbreak with no identified source is not an excuse to wait - it was time to start asking questions

So, the same day, I wrote to our Director of Produce, Miguel Herrera, and looped in our President and VP. I laid out what we knew — the timeline, the states, the case counts — and what we didn’t: whether the implicated supplier was anywhere in our own network.

I proposed a short joint investigation: pull our active suppliers providing Mexican-grown iceberg lettuce or blends containing it, cross-check receiving records against the outbreak’s onset window, reach out directly to any suppliers that matched the profile, and flag anything ambiguous so we could decide together — before any formal recall notice — whether to place a hold on any of our products.

I didn’t have any indication at that point that this touched our supply chain. That was exactly why I wanted to look. Given the volume of produce we move, and what’s at stake — public health first, our reputation right behind it — I’d rather ask the question early and be wrong than wait for someone else to answer it for me.

Public health is at the heart of every food safety manager’s role

Miguel got back to me the next morning. He’d already gone to Topco [a US food-purchasing cooperative] to ask directly whether any product we purchased through Dole or Braga was implicated in the case. The answer was no.

He confirmed that CDC’s findings pointed to iceberg lettuce from Taylor Farms Mexico; a supplier we don’t source iceberg from at all. He was already planning to request an official statement, so we’d have something concrete to share internally and with retail. And he told me they’d keep pressing Topco for confirmation daily, not just once.

That exchange took less than 24 hours. It didn’t take a recall notice, an official supplier name, or a directive from anywhere above us. It took an organisation that already had the right relationships in place, and a food safety team wanting to ask an uncomfortable question before there was any evidence we needed to.

We kept watching the outbreak as it continued to unfold in the news over the following weeks — including a strange twist where FDA walked back one of the implicated lettuce samples as a false positive, even while keeping Taylor Farms epidemiologically implicated. None of that changed our answer. We already knew where we stood, because we hadn’t waited to find out.

I manage food safety compliance across a very large supplier network, and I’ve come to think of outbreak monitoring as a daily habit rather than a reactive task. Part of that is professional responsibility — I don’t want to be the reason a preventable exposure reaches our retailers or their customers.

Part of it is simpler than that: as a food safety leader, I don’t want to be a bad example for anyone watching how this work is supposed to be done. Protecting the company and protecting the public aren’t two different goals in this job. They’re the same one.

Takeaways for food safety professionals

If there’s a single lesson from that week, it’s this: the moment to start investigating isn’t when the recall notices land in your inbox. It’s the moment you see the headline. Think of outbreak monitoring as a daily habit rather than a reactive task.

By the time an official notice arrives, the distributors who are already ahead of it aren’t scrambling — they’re just confirming what they’ve already been watching.

Santoshi Muriki is the Food Safety & Quality Assurance Manager at Affiliated Foods, Inc. (AFI), a wholesale grocery cooperative based in Texas, where she leads compliance efforts across a network of 1,000+ suppliers. She holds an MS in Food Science & Nutrition. Connect with her on LinkedIn.

How to mitigate food fraud when you use co-manufacturers

Food fraud prevention is hard enough for ‘ordinary’ food manufacturers. But for companies that use co-manufacturers, it’s doubly fraught. This article was inspired by real conversations with food safety professionals whose job is to ensure the food carrying their company’s brand name is safe and fraud-free, even though it’s made by someone else.

Food companies that outsource their manufacturing face special challenges when it comes to food fraud prevention. They’re responsible for the integrity of their products but don’t control the environment they’re produced in, the ingredients that are used or the people handling them.

It goes without saying that a lot of trust is required for a successful co-manufacturing relationship. But while “trust but verify” is a popular catch phrase in fraud prevention circles, the verification part is not so easy when it comes to food fraud in these scenarios.

“Trust but verify” — easy to talk about, hard to do

Food brand owners that use co-manufacturers are doubly vulnerable to food fraud, exposed on two fronts:

(1) The co-manufacturer could be an unwitting victim of food fraud.

For example, the co-manufacturer could inadvertently purchase fraud-affected ingredients and raw materials, causing a problem in the finished product, or it could allow finished product or packaging materials to be stolen, creating opportunities for bad actors to resell products and fakes through unauthorised channels.

(2) The co-manufacturer could intentionally defraud or mislead the brand owner.

For example, it could produce more product than contracted and sell the excess through unauthorised channels (‘overruns’); deliberately use lower quality ingredients than contracted; or falsify chain of custody records related to provenance or organic status.

Whether or not the fraud is passed through the supply chain by the co-manufacturer, or is perpetrated by the co-manufacturer, the resulting fraud-affected foods pose risks to brands, their customers and consumers.

How to mitigate these risks

Here is a list of actions food companies can take to reduce the likelihood of food fraud in their supply chains when using co-manufacturers.

Robust supplier approval processes and contracts can ensure that co-manufacturers with a history of unethical practices are not contracted to begin with. Robust purchasing contracts will not prevent fraud but can provide leverage and reduce financial losses if a food fraud problem arises.

On-site audits give you the chance to look specifically for food fraud signals, asking questions such as: Are the co-manufacturer’s raw material records consistent? Do their supplier relationships make sense? Is there evidence of unusual purchasing patterns? Do the raw materials in the warehouse match the invoices you see in the office?

With these questions, you’re gauging whether a supplier could commit fraud even if you haven’t caught them doing it yet.

Example: Your pizza carries the label “Product of Canada”, and your specification requires the co-manufacturer to use only Canadian cheese, but when you visit their site, all the cheese in the cold store carries the name of a US-based dairy company.

Pay particular attention to: who approves new suppliers, how ingredients are received and labelled, the security of finished products and printed packaging materials, and how the co-manufacturer verifies certifications and claims for incoming ingredients (e.g. kosher, organic).

Mass balance is particularly valuable when using co-manufacturers because you’re not physically present on the floor watching production every week. Mass balance is about checking that what goes in roughly matches what comes out.

If the numbers don’t make sense, for example, if you seem to be getting ‘too much’ finished product from ‘too little’ ingredient, or vice versa, it’s a red flag that something may have been swapped, diluted, diverted or mis‑reported.

🍏 How to Perform a Mass Balance: Step‑by‑Step Guide + Downloadable Worksheets 🍏

Supply chain mapping provides visibility about where the co-manufacturer is sourcing their ingredients. Food fraud often happens two or three steps back in the chain — for example, a broker substitutes an ingredient that your co-manufacturer then passes on to you unknowingly (or knowingly).

Work with the co-manufacturer to map the supply chain of your most critical ingredients. This will reveal vulnerabilities within ingredients, such as materials from countries with known food fraud risks, and can be used to assess the robustness of the manufacturer’s food fraud prevention program.

Review their food fraud program. Your co-manufacturer is likely producing for dozens of brands and handling hundreds, if not thousands, of ingredients. By reviewing their food fraud program, you can see if they’ve been diligent about assessing the risks associated with your product.

Questions to ask: have genuine vulnerabilities been identified or glossed over? Is the co-manufacturer relying on suppliers’ CoAs (certificates of analysis) that can easily be forged or faked for food-fraud-vulnerable ingredients? Do mitigation activities specifically address the identified vulnerabilities, or are they simply a set of generic controls?

Verification of supplier certificates is a simple way to check that your co-manufacturer is not being lied to by their suppliers. Ingredients like organic, sustainably sourced, or with specific geographic origin claims are high-fraud-risk because they command premium prices, so rather than accepting a certificate at face value, check its validity directly with the certifying body.

Authenticity clauses and change notifications in contracts are an important tool for brand owners, because without them the co-manufacturer can legally make changes that affect the integrity of your product without telling you, such as swapping a locally sourced ingredient with a short supply chain for a cheaper version with a longer supply chain from a high-risk country.

Authenticity clauses require your co-manufacturer to tell you if there is an authenticity problem with any ingredient, allowing you to make an informed decision about brand risks before the product is made or shipped.

Change notifications require the co-manufacturer to seek approval before they swap an ingredient source, change a formulation, or outsource part of their process. This closes a loophole where unexpected changes to products slip through because “nothing in the contract said we couldn’t.”

Controls for overrun prevention are important with co-manufacturers because overrun fraud is easy to conceal in a shared facility where multiple brands are being produced.

Overrun is when more product gets made than is recorded. The extra product gets sold through channels you didn’t authorise, meaning you lose control of your brand’s distribution, miss out on profits generated from those sales and potentially face consumer complaints if the product is mishandled during distribution.

Overrun frauds are particularly difficult to detect if your co-manufacturer is in a different country, since unauthorised sales are less likely to come to your attention in foreign markets.

Overrun is when more product gets made than is recorded

Mass balance exercises can reveal overruns: unexplained losses could actually be undeclared product leaving the facility. Strict financial penalty clauses in contracts can provide a disincentive for overruns.

Review production, batch and inventory reports during audits: choose a defined period and reconcile them with your own orders and shipments, looking for unexplained production hours, batches, or material usage.

Periodically check marketplace channels and downstream distributors against your own shipment records; product showing up in unusual places or with unexpected lot codes is a red flag.

Takeaways for food professionals

Co-manufactured products pose unique food fraud risks to brands that outsource production, including risks the co-manufacturer may not be aware of, but simply pass through their supply chains, and fraud perpetrated by the co-manufacturer themselves.

Your mitigation strategies should address both types of risk, and should include on-site audits, mass balance exercises, and robust supplier approval processes and contracts with change notification and authenticity clauses.

Learn more:

Common Problems with Contract Manufacturers and How to Solve Them | Amrep Supplier Services

AI and food safety

We all know that popular AI tools like ChatGPT, Meta AI, and Microsoft Copilot can provide misleading answers. When it comes to food safety and science more generally, incorrect answers pose risks to consumers, and even to food companies if they are relied on for food safety plans and hazard analyses.

What’s been missing until now is peer-reviewed documented proof of the unreliability of food safety information provided by such tools.

In a recent paper published in the Journal of Food Protection, researchers posed 20 questions about food handling to four chatbots: Microsoft Copilot, Google Bard, Meta AI, and ChatGPT-3.5, and compared the outputs with the expert opinions of food safety specialists.

Scenarios included typical consumer questions on topics such as expired cream, raw chicken, cross-contamination, swollen cans and unrefrigerated drinks.

In addition to testing whether the chatbot responses aligned with food safety specialists, the researchers also tested whether a change to the way a question was phrased (the prompt) resulted in different risk classifications.

Responses did not align well with expert knowledge and were not consistently reproducible: asking the same question again, or making relatively small prompt changes, could alter the risk classification and advice.

Key takeaway: AI tools are unreliable sources of food safety information — and now we have peer-reviewed evidence.

Source:

Correia, L. and Feng, Y. (2026). Can Consumers Rely on AI Chatbots for Food Safety Advice? A Comparative Analysis of Chatbot Responses and Food Safety Specialists’ Guidance. Journal of Food Protection, [online] 89(9), p.100852. doi:10.1016/j.jfp.2026.100852.

Below for paying subscribers: Food fraud news, horizon scanning and incident reports

📌 Food Fraud News 📌

In this week’s food fraud news:

📌 Method: determining adulteration-level-quantities of other species in buffalo milk
📌 Methanol drink warning for Australia
📌 Warnings for avocados, Basmati rice and fresh vegetables
📌 ‘Food grade’ rose water made with industrial chemicals

Plus, arrests for illegal meat processing, questions about ‘cage-free’ eggs, almond butter concerns, Scottish salmon fraud and more.

The Rotten Apple