‘This in-depth analysis explains the statistical nature of generativeAI and how training on copyright-protected data results inpersistent functional dependencies with respect to the useddata. It highlights the challenges of attribution and noveltydetection in these high-dimensional models, emphasising thelimitations of current methodologies. The study providestechnical recommendations for traceability and outputassessment mechanisms.’
Link: https://www.europarl.europa.eu/RegData/etudes/BRIE/2025/776529/IUST_BRI(2025)776529_EN.pdf
AI Update 