January 25-28, 2027
Bocconi University, Milan, Italy
https://bidsa.unibocconi.eu/cryptography-and-machine-learning-2027-winter-school
AI systems are now deployed at scale, yet we have very few rigorous guarantees about their security, privacy and safety.
For decades, cryptography has developed the tools to reason formally about adversarial behaviour and to prove security against it. Many pressing questions about modern AI are, at heart, cryptographic: making AI-generated content traceable, defending models against adversarial attacks, planting and detecting backdoors, computing on private models and data, and making AI systems safe. This four-day school at Bocconi University, January 25-28, 2027, introduces researchers in cryptography, theoretical computer science, machine learning and AI safety to this fast-growing field, from foundational concepts to the state of the art. Lectures cover secure computation over models, privacy and unlearning, watermarking and steganography, adversarial machine learning, robustness and backdoors, and cryptography for AI safety and agentic systems, among others. The school is open to students and researchers at all career stages.