Can Artificial Intelligence Make Food Safer?

The short answer to this question is: “yes, it can”. There are two major aspects where artificial intelligence (AI) can significantly contribute to making our food safer. On the one hand, it is the improvement of traceability; on the other hand, it is the ability to make better sense of food testing data.

With all the traceability systems that we already have in place and all the certification schemes that have traceability as an integral component, we are still seeing a significant number of recalls for products that are sold to consumers but are not complying with regulations for contaminant limits like heavy metals, pesticide residues or disinfectants.

One (in-)famous example is ethylene oxide, which has kept the food industry busy for more than two years. The substance is often used to reduce salmonella contamination. But using ethylene oxide as a disinfectant for food is not permitted.   Between January 2020 and today, the European System for Rapid Alerts for Food and Feed (RASFF) has seen no less than 861 notifications, the majority of them categorised as “serious”. [NOTE: EC has now set a limit of 0.1 mg/kg for some additives, enforced as of September 1^st ]  The products originated from various European countries, but even the highest count among those with 129 notifications of products originating from France was outnumbered by the 314 products originating from India.

And with all the existing traceability systems and quality control systems in place, one would have thought that eliminating the issue was only a question of months. However, we are still dealing with cases; the last one is a little over a week old (16^th August 2022), where ethylene oxide was found in ground marjoram from Egypt.

So how can AI-based traceability systems help? It is the predictive power that makes those systems so attractive. Being able to predict where affected products might come from and where they are being shipped to. Deep learning enables such systems to identify patterns humans would not easily be able to. Commonalities normally do not stand out but become important in the context.

An example a friend of mine, a criminologist, gave me was that surveillance cameras of three different areas where robberies took place had picked up that one car was spotted in all three areas. Not in the streets where the thefts occurred but two or three roads down from that. Which person would have made the connection that this could be the robbers’ car?

So, improvements in the traceability of products deploying AI can certainly help identify the origins and routes of contaminates or illegal products.  This will help withdraw products from the market faster and likely save consumers’ lives.

The second aspect is the manufacturing on-site AI-supported food analysis to assess rapidly if the incoming raw materials comply with specifications and if they contain adulterants like melamine. Most of us will remember 2008 when the first cases of melamine adulteration in infant formula were reported. The adulterant, especially in combination with the by-product cyanuric acid, is highly toxic and led to hundreds of thousands of infants being hospitalised.  Putting this in context with today, 2008 was also the height of the world food price crisis, causing riots that led to the Arab Spring movement. Today, we are in a similar situation: increasing energy costs, food prices and steeply rising inflation. In addition, the war in Ukraine resulted in significant shortages of sunflower oils and wheat. This is when one would expect to see melamine returning since its addition increases the apparent protein content. And like milk, wheat prices also depend on the content of proteins.

Sending samples to the laboratory will take time and is quite costly. But the novel, AI-powered devices that can be used at the food manufacturing site and operated by factory staff can, within seconds and often without sample preparation, predict if adulterants like melamine or contaminants are present, thereby contributing to food safety for consumers.

Those two developments will, in the mid-term, contribute to making our food safer.

The European Commission has, as part of its FARM-TO-FORK strategy under H2030, funded several projects that develop such tools for both improved traceability and predictive food analytics. As a partner in one of these EU-funded projects starting next month, we look forward to working with the other partners towards this goal.