The integration of collaborative robots (cobots) in industrial environments has transformed human-robot collaboration (HRC), enabling seamless interaction and increased efficiency. Unlike traditional industrial robots, cobots are designed to operate alongside human workers, enhancing flexibility and adaptability in production. However, while physical safety in HRC has been well-regulated through standards such as ISO 10218-2 (2011) and ISO/TS 15066 (2016), the cognitive impact of these interactions remains largely unexplored. The current safety guidelines primarily focus on limiting force, speed and proximity, but fail to address the effects of mental workload, stress, and fatigue, which can significantly impact operator performance and safety. This study investigates the potential of Machine Learning (ML) solutions for the real-time assessment and monitoring of mental workload in HRC. By using physiological data, such as cardiac activity, brain activity or motion patterns, intelligent systems can dynamically adapt robotic behavior to align with the worker’s workload level. This adaptive approach not only enhances safety and efficiency but also aligns with Industry 5.0 principles, emphasizing human-centric automation. The present research provides an overview of state-of-the-art ML models, and adaptive robot strategies that have been explored for safer HRC in industrial settings.