Thanks to its specificities, AI is a versatile tool that can be used in both agricultural and food sectors, basically “from farm to fork”.

Actually, AI, with a learning adaptivity and high interactivity, offers a large autonomy in making decisions and performing tasks autonomously. Thanks to its machine learning systems, AI can adapt to new information and changing conditions, and most importantly, it has the ability to quickly find a specific information in a large amount of data.

So far, the legal challenges surrounding the use of AI are not really new. In fact, when it comes to incorrect outcomes or flawed decisions, the issue revolves around responsibility, primarily involving contractual obligations (2).

For instance, the specific AI characteristics are very useful in precision agriculture, where AI is able to analyse data from multiple sources, such as weather or satellite imagery to offer farmers actionable insights on crop health, irrigation needs, and pest management (3), as well as reducing chemical herbicides (4).

More interestingly, it can also be used in the food manufacturing industry, with quality control, contamination detection, or in supply chain optimization. For instance, ImpactVision created an AI that uses hyperspectral imaging to assess food quality and detect contaminants in real-time. Their technology is used in food production facilities to monitor parameters like ripeness, freshness, and internal defects of food products without damaging them.

But with or without the use of AI, the Food Business Operator remains fully and solely responsible for the safety of its food products placed on the market (5).

With always more precise, more technical and more innovative applications, AI is now increasingly used in food innovation. But the use of AI being not a processing process as such, there are no specific legal challenges directly and specifically related to it. However, AI is often combined with other innovative processes, which, most likely, would lead the final product to a novel food status.

Protein selection or enzyme production are good examples. For instance, with protein selection, as AI is much more than a simple learning machine, it can analyze all proteins faster than anybody/anything else. Thus, it can help alternative proteins producers (plant-based meat or dairy analogues) find the right protein by analyzing protein sequences and select the one with the right characteristics to fulfil some

nutritional, coloring, flavouring or even textural aspects of these alternatives (6, 7). This process is often combined with precision fermentation, which brings a wide perspective of tailor-made products (8).

AI can also be used as sensors, which are also called “e-nose” and “e-tongue”. These tools can be used for sampling classification, detection, or quality control on any foods and beverages, or even to help with flavour and ingredient pairing (9).

In conclusion, the use of AI can be an interesting way to accelerate innovation, meanwhile the probability of falling within the scope of the Novel Food Regulation (10) might be correlated.