Blog

Artificial Intelligence and the Risk Assessment Document (DVR): Requirements and Risks in 2026

Artificial intelligence in the workplace is changing not only the tools companies use but also the way activities are organised and performed. Systems that assign tasks, influence production pace or support operational decisions can have a tangible impact on working conditions and people’s safety.

This transformation creates new opportunities, but it also calls for a careful assessment of the potential consequences for workers’ health and safety.

Recent regulatory developments, particularly Legislative Decree No. 179/2026, make it even more important to understand when the use of AI must be considered as part of a company’s occupational risk assessment.

Which risks need to be examined? When is it necessary to update the Documento di Valutazione dei Rischi (DVR), the Italian Risk Assessment Document? And what measures should companies adopt to integrate these technologies responsibly into their operations?

Artificial intelligence in the workplace: what are the implications for safety?

Introducing artificial intelligence into a company does not, in itself, constitute a new source of danger. Its effects on health and safety depend on the characteristics of the system, the activities in which it is used and the changes it brings about in the organisation of work.

A digital assistant used to summarise documents does not necessarily have the same implications as an algorithm that sets operational priorities in a production department or a system that affects how a machine operates.

This distinction is essential. In some cases, AI can improve the management of tasks, reduce repetitive work or help identify hazardous situations at an early stage. In others, it may affect work pace, decision-making autonomy, cognitive workload and interactions between workers and technology.

For occupational risk prevention, it is therefore important to understand not only which technology is being introduced, but above all how the actual work changes.

The European Agency for Safety and Health at Work (EU-OSHA) points out that AI-based worker management systems can bring organisational benefits but may also affect autonomy, stress and psychological and physical well-being, depending on how they are used.

The assessment should therefore focus on the actual effects on job roles and working conditions, avoiding both the assumption that every AI application is dangerous and the idea that technological innovation is automatically risk-free.

What does Legislative Decree No. 179/2026 provide for workplace safety?

Italian Legislative Decree No. 179 of 7 October 2026, published in the Gazzetta Ufficiale (Italian Official Gazette) on 8 October 2026 with an effective date of 23 October 2026, introduces specific provisions concerning artificial intelligence in the workplace.

In particular, Article 42, entitled “Artificial Intelligence, Health and Safety in the Workplace”, establishes that the use of AI systems affecting work organisation, production pace, how tasks are performed or decision-making processes relevant to safety must be considered as part of the risk assessment required by Article 28 of Legislative Decree No. 81/2008.

The provision clarifies an important principle: when artificial intelligence influences aspects relevant to workers’ health and safety, its effects must be included in the company’s occupational risk assessment process.

Article 42(3) also requires employers to provide workers with appropriate information and training on the specific risks associated with the use of AI systems and on the prevention and protection measures adopted.

However, the legislation does not focus solely on potential concerns. Article 42(4) also promotes the responsible use of artificial intelligence as a tool for preventing workplace accidents and occupational diseases.

These provisions form part of the broader framework already established by Legislative Decree No. 81/2008, Italian Law No. 132/2025 and Regulation (EU) 2024/1689, commonly known as the AI Act.

These legal instruments have different purposes and scopes of application. In particular, the AI Act classifies AI systems according to specific criteria and includes certain applications in employment and worker management among those potentially classified as high-risk. Not every business use of AI automatically falls into this category.

For employers, the first step is therefore to identify the systems actually in use and understand their effects on working conditions, before determining which obligations apply to the specific circumstances.

What risks can arise from the use of artificial intelligence?

The implications of AI for safety can vary considerably depending on the industry, the tasks involved and the degree of autonomy assigned to the systems.

Consider a few applications that may be relevant to businesses.

Algorithms that assign tasks and influence production pace. A system may automatically allocate tasks among operators, change priorities or suggest completion times based on available data. These features can improve efficiency and reduce imbalances in workloads. However, if they are not properly designed and managed, they may contribute to a faster work pace, fewer breaks or increased fatigue. Companies should therefore examine whether the new organisational arrangements affect risks already assessed, including ergonomic and psychosocial risks.

Employee performance monitoring systems. Some applications continuously collect and analyse data on employees’ activities. These tools may help identify inefficiencies or critical situations, but they can also create pressure and reduce workers’ perceived autonomy. Where monitoring affects work organisation and working conditions, its possible health implications must be considered. Separately, companies must also examine the applicable provisions on personal data protection and the monitoring of workers.

AI used for operational instructions and decisions. An AI assistant may suggest procedures, operational sequences or responses to anomalies. This support can be useful, but the information generated may be incomplete, outdated or unsuitable for the actual working conditions. Where such information concerns safety-critical operations, appropriate arrangements for verification, validation and human oversight must be established.

Artificial intelligence integrated into machinery and production processes. Advanced systems can help automate certain operations, monitor parameters or detect anomalies. Introducing these capabilities calls for attention to interactions between operators and machinery, manual intervention procedures and the management of unexpected conditions. The technical and safety requirements applicable to the equipment and processes involved must also be taken into account.

These examples show why it is misleading to speak of a single, undifferentiated “artificial intelligence risk”.

An assessment must start with the specific application, taking account of changes to working activities, the workers involved and the effectiveness of existing preventive measures.

Artificial intelligence and the DVR: when must the risk assessment be updated?

Does introducing an artificial intelligence system always create an obligation to update the DVR? Not necessarily.

It is important to distinguish between considering the effects of AI within the risk assessment and the conditions that make a revision of the Italian Risk Assessment Document (Documento di Valutazione dei Rischi, DVR) mandatory.

Article 42 of Legislative Decree No. 179/2026 establishes that uses of AI affecting the aspects of work identified by the provision must be assessed under Article 28 of Legislative Decree No. 81/2008.

Article 29(3) of Legislative Decree No. 81/2008, by contrast, specifies the circumstances in which the risk assessment must be revised, including changes to production processes or work organisation that are significant for workers’ health and safety.

When these conditions arise, the assessment must be revised immediately and preventive measures must be updated. The DVR must be revised within thirty days of the event triggering the requirement, without prejudice to the obligation to document updates to preventive measures immediately and communicate them to the workers’ safety representative (Rappresentante dei Lavoratori per la Sicurezza, RLS).

Consider two different situations.

A company introduces an AI assistant solely to support administrative activities, without significantly changing job roles, work organisation or safety conditions. In this case, the use of the technology does not automatically require the DVR to be revised, although any relevant effects must still be examined.

By contrast, a company adopts an algorithmic system that changes the distribution of workloads, increases the frequency of operations or influences decisions affecting the safety of a process. If significant changes to health and safety arise, the risk assessment must be revised and the DVR updated in accordance with the applicable requirements.

The decisive factor is therefore not the mere presence of AI, but the significance of the changes it introduces in relation to occupational risks.

Responsibility for the risk assessment remains with the employer under Article 17 of Legislative Decree No. 81/2008, with the involvement of the occupational health and safety roles required by law.

For further information on the role of this document in company management, see Safetyone’s article on the Risk Assessment Document (DVR).

This issue is distinct from the use of AI to generate technical safety documentation, which is addressed in Safetyone’s article Artificial Intelligence and Workplace Safety: Risks, Limitations and Responsibilities.

Information, training and prevention: what must companies ensure?

Risk assessment is the starting point, but the safe use of artificial intelligence also requires appropriate organisational measures and suitable competencies.

Article 42(3) of Legislative Decree No. 179/2026 expressly addresses the provision of information and training to workers concerning the specific risks connected with AI systems and the prevention and protection measures adopted.

The content must be tailored to the applications actually used and the activities performed. A worker using an automated decision-support system, for example, needs to understand the relevant operating instructions, the system’s limitations and the procedures for handling any anomalies.

Alongside legal obligations, it is good practice to clearly define who may authorise certain uses, which decisions require human verification and how unexpected system behaviour should be reported.

Involving HSE functions, operational managers and, where necessary, IT, HR and legal specialists can make the integration of new technologies into company processes more effective.

AI can also create new opportunities for prevention, from identifying unusual situations and analysing data to supporting the planning of inspection and maintenance activities.

These benefits must nevertheless be assessed in the relevant context. Technology can support preventive decisions, but it does not replace the responsibilities and checks assigned to competent professionals.

How can companies assess the impact of AI on workplace safety?

A sound approach to introducing artificial intelligence is to adopt a review method proportionate to the characteristics of the applications in use.

The first step is to identify the systems used in company processes, establish which activities they influence and understand whether their use changes working conditions.

This analysis makes it possible to determine which further investigations are needed, which functions should be involved and whether existing preventive measures remain adequate.

Five questions can help a company carry out an initial review:

Which AI systems are actually being used? Their functions, purposes and the processes in which they operate should be identified.
Does AI change job duties, production pace or operating procedures? Any changes that could be significant for health and safety must be identified.
Which decisions are made or suggested by the system? It is necessary to understand which activities require checks, validation or human intervention.
Are workers aware of the specific risks and preventive measures? Information and training must be appropriate to the actual conditions of use.
Are the risk assessment and preventive measures still adequate? The effects of changes must be examined and, when the conditions prescribed by law are met, the necessary revision must be carried out.

These questions do not replace a technical assessment, but they can help identify matters requiring further examination.

Artificial intelligence offers businesses significant opportunities for innovation. To realise these benefits sustainably, however, new technologies must be integrated into an occupational risk prevention system that can respond to changes in activities and working conditions.

The real question is not simply how much AI can improve work, but how to ensure that this change takes place while protecting people’s health and safety.

Safetyone Ingegneria supports companies in analysing occupational risks, assessing organisational changes and managing occupational health and safety obligations.

To explore the possible implications of introducing new technologies for your company’s occupational risks, you can contact Safetyone Ingegneria.