ITALY Trends and Developments Contributed by: Michela Bani, Alessandro Paone and Giacomo Bertelli, NIUS Legal and HR Solutions
IA, remote controls and Article 4 of the Workers’ Statute The adoption of AI systems in corporate investigations is often located in a border area with the discipline of remote controls. Even when AI is integrated into work tools (computers, software, digital platforms), its abil - ity to systematically and continuously monitor requires a restrictive reading of the exceptions provided for by Article 4, paragraph 2, of the Workers’ Statute. The doctrine and the supervisory authorities have clarified that the installation of AI systems suitable for allowing control, even indirectly, over the activity of workers requires: • the prior trade union agreement or administrative authorisation; • the provision of specific and detailed information; • a data protection impact assessment (DPIA); and • the strict application of the principles of propor - tionality and minimisation. In the context of internal investigations, the use of AI cannot therefore result in generalised or exploratory surveillance, but must be limited to specific hypoth - eses, based on a concrete and documentable suspi - cion of wrongdoing. Risk assessment, governance and accountability The use of AI in internal investigations requires the adoption of a structured governance system. In par - ticular, the employer must be able to demonstrate that they: • have assessed in advance the legal, organisational and discriminatory risks related to the use of AI; • have defined clear roles and responsibilities in the management of the systems; • have implemented auditing and continuous moni - toring mechanisms; and • are able to explain and justify the decisions taken on the basis of algorithmic outputs. The principle of accountability, already central to the GDPR, is further strengthened by the legislation on AI: the responsibility remains with the person using the system, regardless of the degree of technical auton - omy of the algorithm.
Trade union profiles and collective bargaining The growing pervasiveness of AI makes collective bar - gaining a privileged tool of regulation. Recent trade union agreements show how it is possible to regulate the use of AI in analysis, monitoring and decision- making support activities, introducing additional guar - antees in terms of voluntariness, anonymisation, time limits for data storage and equal observatories. In the context of internal investigations, the involve - ment of trade union representatives may constitute not only a formal fulfilment, but also a factor of sub - stantial legitimacy of the entire investigative system. Application risks: bias, opacity and reversal of the burden of proof The use of AI in internal investigations poses signifi - cant risks, including: • algorithmic bias, with possible discriminatory effects; • opacity of decision-making processes (black box effect); and • difficulty in contesting the results by the worker. The violation of the obligations of transparency and explainability can determine, in court, a substantial reversal of the burden of proof, giving the employer the task of demonstrating the legitimacy and non- discrimination of the decisions taken. In this scenario, AI represents not only an opportunity for investigative efficiency, but also a legal risk factor that must be governed with appropriate regulatory, organisational and cultural tools. AI governance and safekeeping: conclusions The use of AI in internal company investigations today represents a strategic junction in risk management, in the structuring of compliance systems and in strength - ening the organisational responsibility of the com - pany. AI makes it possible to overcome fragmented and reactive investigative models, offering advanced tools for the analysis, traceability and oversight of pro - cesses, capable of increasing the effectiveness and timeliness of employer action.
257 CHAMBERS.COM
Powered by FlippingBook