Home Center: Research Center AI4Society

The rapid spread of generative artificial intelligence systems has transformed how organizations produce and use information, opening up new opportunities but also introducing significant risks. Content generated by large language models (LLMs) can, in fact, appear plausible even though it's unfounded (“hallucinations”), producing information that is incorrect, opaque, and potentially harmful.

This content, referred to as unsafe, poses a risk not only to the reliability of digital services but also to the social responsibility of organizations toward their employees, customers, and stakeholders. As businesses and public administrations increasingly adopt AI solutions, it becomes necessary to move beyond traditional approaches to cyber risk management. This is to ensure that the information produced by AI systems is “cyber safe”—accurate, protected, and not harmful to society. The theoretical distinction between information security and safety—borrowed from the fields of functional safety—means rethinking data governance practices to encompass not only protection from intentional threats (security) but also the prevention of accidental harm (safety). In this context, there is a need to assess the strategic impact of AI solutions and, prior to system design, define the control requirements for infrastructure, applications, processes, and data that can mitigate security and safety risks to ensure the reliable use of AI in organizational settings.
 

In line with this need, this project aims to launch a “Cyber Safe AI Innovation” laboratory (XAI Lab) within the AI4Society research center at Luiss, in close collaboration with the Cyber 4.0 Competence Center, of which the university is a member. The lab was created to support small and medium-sized businesses, large corporations, and public administrations in developing reliable, secure, and responsibly integrated artificial intelligence solutions for their organizational processes. Inspired by similar initiatives at international universities, the lab adopts an Action Design Research (ADR) approach to guide organizations in assessing critical issues related to the strategic use of AI, co-designing AI prototypes in controlled environments, and leveraging the results for technology transfer. The lab views security and safety not as isolated technical requirements, but as broader objectives to be defined and governed through collaborative models that foster shared responsibility and cooperation between public and private stakeholders. The lab will serve as a hub for the growth of a community of practice capable of addressing issues related to secure digital innovation in an interdisciplinary manner, including through stakeholder engagement and capacity-building events at the Luiss locations in Viale Romania, Viale Pola, Villa Blanc, and Milan. The lab will organize workshops, seminars, networking events, and joint research and training activities involving students, faculty, researchers, and experts from the Luiss and Cyber 4.0 networks.