Abstract

The prevailing discourse regarding AI systems in the workplace predominantly focuses on opportunities to enhance occupational safety and health. However, there is insufficient mapping of the negative implications these technologies may have on the workforce. This paper examines the health impacts of AI, specifically focusing on mental and social well-being, referred to as the workplace psychosocial environment. It argues that addressing psychosocial risks (PSRs) requires an integrated approach across labour, equality, and data protection laws.

About the authors

Tahmina Karimova is a lawyer and Law Research Specialist at the Research Department of the International Labour Organization (ILO). Her expertise includes public international law, sustainable development, and international labour standards. She has previously worked as a Human Rights Officer for the United Nations and holds a PhD in International Law from the University of Geneva.

Introduction

The world of work is undergoing significant transformations due to digitalisation and AI-based technologies. AI is increasingly used throughout the entire employment cycle, including recruitment, training, monitoring, remuneration, and dismissal. This shift impacts how work is designed and organised, leading to profound consequences for worker health. While current attention is on technological benefits, there is little systematised knowledge regarding health impacts, particularly PSRs, which are defined as factors in work design or management that increase work-related stress.

Setting the scene

AI-based technologies include advanced robotics, AI-based algorithmic management (AM), and smart digital systems. AM involves the delegation of managerial functions to algorithms that collect and process workspace data to inform or automate decisions. Research indicates that these tools are becoming standard; a survey of 6,047 mid-level managers in six high-income countries found that 74 per cent of firms use at least one AM tool to monitor or evaluate employees. Surveillance capabilities are associated with adverse effects; a US survey of 1,273 workers revealed that 46 per cent of those monitored ‘all the time’ felt they worked too fast, and 53 per cent reported feeling anxious at work.

Psychosocial risks and AI systems: Interplay and data availability

Psychosocial factors include job content, workload, job control, and interpersonal relationships. AI systems can ‘augment’ or create new risks such as deskilling, cognitive overload, and work intensification. Intensive surveillance through digital tools like Microsoft Teams can generate vast amounts of activity data, including call participation and chat frequency, which may undermine worker dignity and trust. Furthermore, excessive data collection often lacks transparency, leading to a perceived lack of fairness. High levels of automation have also been found to have a detrimental effect on situational awareness and the feeling of control over one’s tasks.

Regulatory frameworks addressing AI associated PSRs

Current international labour standards, such as ILO Convention No. 155, provide a basis for mental health protection but are technology-neutral. Emerging regulations, such as the EU AI Act, classify the use of AI in employment as high-risk, banning systems that infer emotions in the workplace. National responses vary: Switzerland prohibits monitoring employee ‘behaviour’ through AI, while Australia (NSW) regulates workplace surveillance with mandatory notice requirements. Spain’s ‘Rider Law’ imposes algorithmic transparency, and Portugal has introduced protections against permanent connection for teleworkers. Bulgaria’s Labour Code now requires employers to provide written information about algorithmic management systems and allows workers to request human review of automated decisions.

Conclusion

Tailoring policies for the safe use of AI requires understanding its impact on the psychosocial landscape. The paper identifies a range of emerging factors, including intrusive surveillance, loss of autonomy, and a lack of transparency. It suggests that future regulatory improvements should include new occupational rights, such as the right to information regarding AI systems, the right to human review of decisions, and the ‘right to disconnect’ (R2D) to protect workers’ private lives and mitigate constant digital monitoring.