Applications of artificial intelligence and machine learning in food processing, quality control, and food safety: a critical review

Artificial intelligence (AI), machine learning (ML), and deep learning (DL) are increasingly being applied to support data-driven decision-making in food processing, quality control, and food safety. This review provides a critical assessment of recent technological advances and examines how AI models are being integrated with computer vision, spectroscopy, hyperspectral imaging (HSI), biosensors, Internet of Things (IoT) platforms, and intelligent monitoring systems throughout the food value chain. This review brings together evidence from academic studies and industry-based research to examine how AI is currently used in the food sector. The applications considered include process control and optimization, fermentation management, quality grading, authenticity assessment, shelf-life prediction, contaminant detection, and support for food-safety decisions. Recent research extends beyond conventional prediction toward adaptive decision support and, in a smaller number of cases, partially autonomous food systems. Approaches such as explainable AI, multimodal learning, foundation models, digital twins, intelligent packaging, and agentic AI are being investigated for real-time monitoring, data integration, and production management. Their level of maturity, however, varies considerably, and many applications remain conceptual or confined to laboratory settings. Data heterogeneity, poor transferability, limited external validation, and regulatory uncertainty continue to restrict industrial adoption. Although one industrial-scale drying study reported a 12.8% reduction in energy consumption, external validation of food-quality models has resulted in markedly higher prediction errors. This contrast illustrates the difficulty of transferring performance from controlled datasets to new products or production environments. Rather than examining these technologies separately, this review considers their roles across food processing, quality control, and food safety, together with the evidence for industrial validation, sustainability, and governance. Further progress will depend on whether models remain reliable outside their development datasets, not simply on incremental gains in internally reported accuracy.