About the Workshop
Foundation models, large language models, and agentic systems are rapidly reshaping the privacy landscape of machine learning. Across these paradigms, privacy risks are amplified by opacity: training data, post-training pipelines, alignment procedures, system components, and deployment contexts are often only partially visible to researchers, auditors, and users. These systems can expose sensitive information through memorization, retrieval, long-term memory, tool use, and cross-context information flow. At the same time, their opacity makes privacy assessment difficult. Together, these developments challenge existing definitions, benchmarks, mitigation strategies, governance frameworks, and accountability mechanisms. This workshop will provide a timely forum for examining memorization and privacy risks in opaque AI systems, with emphasis on how these risks should be defined, measured, mitigated, and governed in deployment. It will bring together researchers from privacy, machine learning, HCI, law, and agentic AI around a central question: what should privacy mean when large opaque models are trained, adapted, deployed, and allowed to act on sensitive information?
