With the emergence of AI, there is increasing interest in using predictive models [5]. This provides a number of tools that can assist farmers in monitoring environmental factors such as weather, temperature, water usage, and soil quality [5]. Ensuring food safety is central to achieving sustainable food security [6], which includes the monitoring of hazards, including biological (e.g., bacteria, viruses, and parasites), chemical (e.g., heavy metals, pesticides, and mycotoxins), and physical (e.g., metal particles, glass fragments) [7]. These hazards can pose a risk of contamination and eventually unsafe food production for consumers, with a serious risk to human and animal health [7]. Also, one of the core elements of food safety measures is to forecast and prevent future risks [8, 9]. Therefore, ensuring food safety throughout the entire food chain, by controlling and monitoring food safety, is crucial through early warning systems [10]. Machine learning algorithms can be trained on historical data and build analytical models [11], which can be used to forecast, predict, and prevent future threats to food safety, including outbreaks [8].

Moreover, it is worth noting the potential threat of climate change to food security globally: The risk of hunger, which is one of the impediments to sustainable development cited by the United Nations. Machine learning can play a role in the agricultural supply chain, which includes pre-production, production, processing, and distribution [12].

In recent years, a proportion of the food production industry has introduced AI [13], which has enhanced the digitalization of data during the food production processes, tracing, monitoring, and inspection. Therefore, AI has emerged as a pivotal technology in this regard. Hereby, the digitalized data can be used by other AI algorithms to identify patterns that might be overlooked through a manual analysis, and can potentially predict potential outbreak sources and spread with higher accuracy and speed (although such a process can lead to hallucination with the current versions of AI). In addition, AI models could be trained to match consumption patterns to food exposure from persons who fell ill during an outbreak [14]. Also, AI can be used to compile large qualitative data collected during the outbreak’s interview process [15]. Utilization of these models could facilitate the rapid detection of the source of foodborne contamination, which is crucial for food safety and would lead to less foodborne illness overall [14].

For example, the different machine learning-related models can be trained on historical contamination data to predict the probability of contamination during the process of the supply chain. It includes supervised learning to identify anomalies in production line data or unsupervised learning to detect unusual patterns indicative of potential risks. Such an approach, when combined with real-time monitoring, has the potential to do early detection of contamination, which can lead to more efficient response strategies [5, 13].

Machine learning can be used to increase our ability to predict foodborne illness etiologies, geographical spread, and potential outbreak risks [16]. Models, such as anomaly detection using image recognition, can be used to assess hygiene practices (for example, analysing handwashing techniques, evaluating food safety, and predicting contamination risk); however, these models should be used with caution, as AI systems can have errors in their reading capabilities [17]. Also, supervised learning can link syndromic surveillance to the causative agents—Although the methods for outbreak detection require overcoming the risk of under-reporting and incorrect reporting [18].

Hence, the acceleration in such predictions can help public health agencies in early interventions, which would prevent the population from being exposed to contaminated food and, eventually, prevent larger outbreaks [19].

Spectra-based analysis and hyperspectral imaging have the potential to identify, rapidly (in a few hours instead of 24–72 hours for pathogen detection and strain identification) and with precision, bacteria in environmental samples, which is crucial for food safety, and therefore in outbreak investigations [20]. As well, it can assist in differentiating bacterial strains and understanding their resistance to antibiotics, host specificity, and virulence. Furthermore, machine learning can model how bacteria grow and interact within populations, offering insights that can streamline the time, resources, and expertise needed for comprehensive food safety analysis and decision-making [20].

The rapid collection of image data and advances in imaging technology make AI ready for commercial use and integration into food safety testing systems. It can potentially be adapted to various food types and conditions, contributing to a more sophisticated understanding and control of foodborne pathogens. This integration of AI is essential, given the complex nature of pathogen detection and characterization and the need for a capable workforce. Especially as traditional labor-intensive methods face the challenges of workforce attrition [6], machine learning facilitates the automation of many laboratory processes.

To be able to enhance the utility of AI, the following strategies are needed: i) enhancing data sharing agreements among all stakeholders, using strict data privacy regulatory measures and an ethical framework [21], ii) fostering multi-disciplinary partnerships, and iii) promoting education in data science, software engineering, and design thinking for food safety professionals [16].

Governance considerations: AI is a form of technology meant to enable health systems towards the betterment of health [22]. Although AI is not a fully autonomous system, it has the tendency to make independent decisions, which raises concerns regarding the theme of accountability for the related outcomes [21]. Policy makers regulate AI in a reactive fashion, i.e., instead of generating full laws, they produce framework guidelines that outline the AI-related policies [22]. In general, the European Union (EU) is considered a pioneer in developing a blueprint for the international community on how to proceed with AI regulations [20]. This includes protecting the rights of individuals to the utility of AI, and establishing a standardized definition of an AI system [22]. Additionally, AI tends to be cost-effective when compared to the traditional approach of using a microbiology laboratory facility [20, 23]. Finally, AI-assisted approaches tend to detect bacterial classification at a higher accuracy [20].