CHAPTER I: Di cosa parliamo quando ci riferiamo all'Intelligenza Artificiale?

Presenter: Prof. Fabio Roli - Italian language
Release date: June 2026

First chapter of the InfoAICert training programme: an accessible introduction to modern Artificial Intelligence, designed for people without a technical background. The session starts with the terminology - what we mean by AI today and why, in practice, it is largely equivalent to “machine learning” - and then explores how pattern recognition works, how model performance is measured, and what the open challenges of this technology are.

Topics covered:

• Terminology: what we mean when we talk about AI
• Machine learning for pattern classification
• How patterns are represented in computers
• How machine learning performance is measured
• Historical evolution and current performance
• Open challenges in the use of machine learning

Fabio Roli is Full Professor of Computer Engineering at the University of Genoa and the University of Cagliari, and Director of the sAIfer Lab. He is internationally recognised as one of the pioneers in the field of Artificial Intelligence security. He is included in Stanford University's list of the world's most highly cited researchers.

CHAPTER II: Evoluzione storica e tecnica dell’AI dagli algoritmi agli "AI system" del regolamento europeo

Presenter: Prof. Fabio Roli - Italian language
Release date: June 2026

In this second chapter we trace the evolution of AI from the 1950s to the present day: from rule-based expert systems to machine learning, and finally to transformer-based generative models such as ChatGPT. We examine the role of data, models, and GPUs in recent developments, the structural limitations of modern AI - including bias, fragility, and generalisation - and how all of this relates to the definition of an “AI System” under the European Artificial Intelligence Act.

Topics covered:

• Origins of AI: from the Dartmouth Workshop to the 1980s
• Symbolic AI and expert systems: limitations of the algorithmic approach
• The emergence of machine learning
• Deep learning, neural networks, transformers, and LLMs
• The role of big data and GPUs
• Limitations of modern AI: bias, spurious correlations, and overfitting
• From technology to regulation: what is an “AI System” under the AI Act?

Fabio Roli is Full Professor of Computer Engineering at the University of Genoa and the University of Cagliari, and Director of the sAIfer Lab. He is internationally recognised as one of the pioneers in the field of Artificial Intelligence security. He is included in Stanford University's list of the world's most highly cited researchers.

CHAPTER III: Introduzione alla Trustworthy AI

Presenter: Prof. Fabio Roli - Italian language
Release date: June 2026

In this chapter: when can we really trust an Artificial Intelligence system? Starting from the European principles of Trustworthy AI - reliability, robustness, security, fairness, transparency, and accountability - we explore the main risks associated with AI systems, including bias, lack of robustness, adversarial attacks, and decisions that are difficult to interpret, as well as the methodologies used to make AI systems more trustworthy, in connection with the European AI Act.

Topics covered:

• What is meant by Trustworthy AI and the European principles
• Why modern AI raises trust-related challenges
• Concrete risks: bias, systematic errors, and decisions that are difficult to understand
• Robustness and security: adversarial attacks
• Explainable AI and performance validation
• The connection between Trustworthy AI and the AI Act

Fabio Roli is Full Professor of Computer Engineering at the University of Genoa and the University of Cagliari, and Director of the sAIfer Lab. He is internationally recognised as one of the pioneers in the field of Artificial Intelligence security. He is included in Stanford University's list of the world's most highly cited researchers.

CHAPTER IV: Trustworthy AI: Soluzioni e Strumenti

Presenters: Maura Pintor & Luca Demetrio - Italian language
Release date: June 2026

In this chapter we introduce the fundamental concepts behind attacks on AI systems - with a particular focus on evasion attacks - and how to assess their impact using dedicated tools such as SecML-Torch. We explore the assessment of complex systems (Compound AI Systems) and InfoAICert’s vision of making security methodologies accessible even to non-specialists.

Topics covered:

• Overview of threats to AI algorithms
• Evasion attacks and impact assessment
• From theory to practice: SecML-Torch dashboard demo
• Assessment of complex systems: towards Compound AI Systems
• Limitations of current tools and the future vision of InfoAICert

Maura Pintor is Assistant Professor at the Department of Electrical and Electronic Engineering at the University of Cagliari, where her scientific work focuses on studying the vulnerabilities of intelligent systems and developing robust methodologies for protection against adversarial attacks.

Luca Demetrio is Assistant Professor at DIBRIS, University of Genoa, specialising in cybersecurity, with a particular focus on AI security and adversarial machine learning techniques.

CHAPTER V: Come funzionano davvero i Large Language Models?

Presenter: Antonio Emanuele Cinà & Matteo Todescato - Italian language
Release date: August 2026

In this chapter: a practical look at how models such as ChatGPT, Claude, and Gemini actually work - statistical prediction of text rather than an understanding of the world - and what hallucinations are, with concrete examples, including in the legal and compliance domains. In the second part, we present techniques to mitigate these risks, with a particular focus on RAG and Graph RAG systems, which anchor responses to verifiable sources, even in complex regulatory contexts such as the GDPR, ISO 27001, and NIS2.

Topics covered:

• How LLMs work: the concept of the “most-likely next token”
• The limitations between prediction and semantic understanding
• Hallucinations: definition, causes, and examples (e.g., the Mata v. Avianca case)
• From “memory-based” answers to answers grounded in verifiable sources (RAG)
• From document retrieval to structured knowledge (Graph RAG)
• Examples applied to compliance management

Antonio Emanuele Cinà is an Assistant Professor (RTDA) at the University of Genoa. He received his PhD in Computer Science with honours from Ca’ Foscari University of Venice in 2023 and was a postdoctoral researcher at the CISPA Helmholtz Center in Saarbrücken. His research focuses on the reliability of AI systems in adversarial settings, studying vulnerabilities caused by spurious correlations and developing mitigation techniques. He is a member of the IEEE Computer Society and has been an Associate Editor of the International Journal of Machine Learning and Cybernetics (Springer) since 2025.

Matteo Todescato is a Product Innovation Professional at Tinexta InfoCert, where he has been working since 2018. He obtained a Master’s degree in Computer Science from the University of Padua in 2024, where he was a member of the SPRITZ (Security and Privacy) research group. During his academic career, he contributed to high-level publications and workshops in the field of post-quantum cryptography.

CHAPTER VI: AI regulations

Presenters: Giuliana Marzola - Italian language
Release date: August 2026

Sixth and final session of the InfoAICert training programme: an overview of AI regulations at European and international level, with a focus on the AI Act - the world’s first binding AI regulation, based on a risk-based approach. A comparative framework to understand how regulation affects innovation and global competitiveness, from obligations for high-risk systems to the main international frameworks, including the NIST AI RMF and ISO/IEC 42001 standards.

Topics covered:

• Europe: the AI Act and its application in Italy
• The role of technical standards
• International regulatory approaches (USA, Asia, Latin America)
• Ethical guidelines and codes of conduct
• Future perspectives

Giuliana Marzola, Technical Compliance Expert at Tinexta InfoCert, holds a degree in Mathematics. She began her career as an Assembler and C developer and subsequently became an expert in cryptography. She has participated in standardisation groups at ETSI and, for a brief period, worked as a CISA-certified information security auditor. For several years, she has been working at Tinexta InfoCert in the field of technical compliance, with a particular focus on trust services. She currently participates in the InfoAICert working group, where she contributes her expertise on compliance with regulations and technical standards.