AI Making Inroads in Food Safety Labs
We’ve often written about how artificial intelligence is penetrating the product development process and plant operations. While less reported, the technology is having a quietly growing influence in food safety laboratories – although not in all labs.
Molecular assays, automated enumeration systems and ATP (adenosine triphosphate) bioluminescence have been the standards for testing labs, but even these “rapid microbiological methods” could be hastened. Plus, how rapid is the analysis of all the data generated? And how much more time is spent on decision-making?
The same pressures for speed, accuracy and efficiency that are affecting all aspects of your business are impacting testing labs – whether your company does this testing in-house or contracts it out. Processors need faster results (which enable earlier product release), reduced manual handling (lowering variability and labor demands), improved sensitivity (especially for stressed or low-level organisms) and digital traceability (supporting audit readiness and data integrity).
“Food safety laboratories are undergoing a significant shift as the food industry faces increasing pressure for speed, accuracy and transparency,” Wesam Al-Jeddawi, co-founder and chief scientific officer at Core Catalyst Food Sciences, wrote in a Food Safety Tech article. He also talked to us for this story.
Core Catalyst Food Sciences provides microbiology testing, chemistry testing, and shelf-life studies for food & beverage manufacturers, with a focus on defensible data, rapid turnaround and practical support for quality and safety programs.
“Traditional microbiological methods remain foundational, but they are often too slow to support today’s accelerated production cycles and complex supply chains,” he says. “As a result, laboratories are adopting rapid microbiological methods, digital data systems and artificial intelligence to enhance decision-making and reduce risk.
“These technologies are not replacing scientific expertise,” Al-Jeddawi wrote in Food Safety Tech, “they are expanding what laboratories can deliver. When integrated thoughtfully, they improve efficiency, strengthen data integrity, and help manufacturers identify issues earlier in the production process.”
Once the sampling and testing is done, food safety labs can churn out a lot of data. And they’re spitting out more and more data all the time. But how to interpret it all, especially in near-real-time, much less make automated decisions based on all this data?
Competitors Cooperate for Food Safety
In 2024, BioMérieux and its parent Mérieux NutriSciences launched the Trusted Third Party (TTP) initiative, a collaborative effort with a handful of food processors to anonymously share information on food safety problems.
Barry Callebaut, Danone, Mars, Mondelēz International, Nestlé and PepsiCo were involved in the exploration and design of the TTP model. Each company shared initial data sets to build the proof of concept and evaluate potential interest, helped identify key business questions and food safety issues, contributed to the platform design and now support the model’s development and expansion.
“Data is confidentially and securely aggregated from all participating partners and is combined with other publicly available data of interest (commodity pricing, food safety events, weather, etc.) to deliver consolidated anonymous insights and support business critical decision-making, strengthening partners’ food safety risk management programs and supply chain integrity,” said the initial announcement.
“Say you’re having a food safety issue relating to your seafood products,” says John Shultz, senior director of marketing & sales at BioMerieux. “You might see that other processors who are using shrimp from Vietnam are having the same problem, but not those who are sourcing shrimp from Thailand. You can make an informed decision based on that data.”
BioMérieux and Mérieux NutriSciences are accepting additional members into TTP.
Enter AI
The pursuit of automated interpretation of data is almost as old as the computerized creation of that data. As in other fields, AI can speed the interpretation of all that data and, at the leading edge, can suggest or even make decisions as a result of those interpretations.
Artificial intelligence is beginning to influence how laboratories interpret and manage microbiological data. Al-Jeddawi says its most impactful applications include:
- Identifying patterns across historical testing data.
- Predicting spoilage and contamination risks.
- Automating data checks, reducing transcription errors.
- Supporting root-cause analysis with more complete datasets.
“When paired with a laboratory information management system, AI helps laboratories transition from reactive testing to proactive risk management,” he says. “Instead of simply reporting results, labs can provide insights that help manufacturers prevent issues before they occur.”
BioMerieux, which develops technologies and services used by food labs, last year bought Neoprospecta, a Brazilian biotechnological company that was developing AI services for food safety labs. Subsequently, BioMérieux developed Smartbiome, an AI-assisted software that combines high-precision DNA analysis with advanced bioinformatics tools to understand spoilage issues. By establishing the root causes of non-quality in finished products, processors are able to make informed decisions and implement effective risk prevention measures upstream in the process.
BioMerieux noted around 80% of microbial food safety tests are performed for spoilers and quality indicators. Smartbiome acts in real time and not only detects issues but recommends solutions.
“You stop looking at data as a single point in time, instead as a continuum, where you can see your process heading toward a problem,” says John Shultz. Despite his title of senior director of marketing & sales at BioMerieux, Shultz is a molecular biologist, pushed into a sales & marketing role because of the increasing complexity of the technology, and he sits on the company’s AI committee.
“We are investing in converting information – copious amounts of microbiological data – into decision-making, used in the moment for a process control purpose to generate a go/no-go signal for the operator to decide if they can proceed with some action,” he says. “Did I meet my [food safety] specifications? Is it good enough microbiologically so I can send this into commerce? Or do I need another cleaning?”
Smartbiome lets companies make informed decisions faster and optimizes production processes for better risk management. “Once you’ve done that, you can make better decisions in the moment, you move on, no waiting,” Shultz says.
Furthermore, if you retain that information and begin to connect it to a continuum of events that relate to operational risk and control across the manufacturing plant, you can make decisions based on “the continuous whole of the data that’s been generated,” Shultz continues.
“You can begin to draw trends from that information. Can you connect that microbiological information about trends to other forms of information? To supplier data? To control data in the process, like temperature, hold times, mixing conditions? You can start making decisions you couldn’t four or five years ago. Some of these questions you never could have answered.”
The Smartbiome knowledge base contains extensive information on more than 3,000 microorganisms, detailing optimal growth conditions (pH, temperature), resistance mechanisms (including sanitizers) and their specific impacts on food products – for example, changes in odor or shortened shelf life.
Smartbiome can spot the contaminant affecting the finished product, identify how it got into your operation, suggest interventions to control this spoiler and recommend how to prevent future spoiler contamination.
“We call it augmented diagnostics, augmenting what was a singular diagnostic test with a software tool layer that combines time, continuous monitoring, looking at the process. Connecting all that stuff together is a big data puzzle, and that’s where AI tools can come in – to give visibility to the signals that before were just noise but in reality are real signals of risk.”
It’s fascinating stuff, but all these things have been within the capabilities of humans. Maybe not so fast or as well, but humans have proven capable of building such “expert systems” to monitor signals and suggest responses.
However, with AI, “You no longer need a human to build a model,” says Shultz. “The model builds itself. Now we’re letting the decision making originate from the context of the data itself.
“You’re not having a human codify anything. You’re allowing the system to consider all interrelated data sets . And it may be making decision based on things humans may never have considered before — which is pretty radical.”
Shultz concludes, “We’re at the precipice of a very transformative time in diagnostics and food safety and quality testing.”
About the Author
Dave Fusaro
Editor in Chief
Dave Fusaro has served as editor in chief of Food Processing magazine since 2003. Dave has 30 years experience in food & beverage industry journalism and has won several national ASBPE writing awards for his Food Processing stories. Dave has been interviewed on CNN, quoted in national newspapers and he authored a 200-page market research report on the milk industry. Formerly an award-winning newspaper reporter who specialized in business writing, he holds a BA in journalism from Marquette University. Prior to joining Food Processing, Dave was Editor-In-Chief of Dairy Foods and was Managing Editor of Prepared Foods.