20.500.12556/DKUM-96363 Zagotavljanje skladnosti z GDPR pri obdelavah podatkov z modeli umetne inteligence diplomsko delo visokošolskega študijskega programa Informacijska varnost Ensuring gdpr compliance in data processing using artificial intelligence models Umetna inteligenca (UI) in sistemi strojnega učenja postajajo osrednji del sodobnih digitalnih storitev, pri čemer njihova sposobnost obdelave velikih količin podatkov odpira pomembna vprašanja glede varstva zasebnosti in skladnosti z zakonodajo, kot je Splošna uredba o varstvu podatkov (GDPR). Namen tega dela je analizirati ključne izzive zagotavljanja skladnosti UI sistemov z zahtevami GDPR ter oceniti mehanizme, strategije in pravne okvire, ki jih omogočajo ali zahtevajo obstoječi evropski in mednarodni regulativni dokumenti. UI, še posebej v svojih generativnih in napovednih oblikah, pogosto temelji na množičnem zbiranju in profiliranju podatkov, kar odpira dileme glede zakonitosti obdelave, privolitve ter načela najmanjšega obsega podatkov. EU je uvedla tveganjsko zasnovan pristop, ki v ospredje postavlja zanesljivost in preglednost UI sistemov, kljub temu pa obstajajo še neodgovorjena vprašanja o razmerjih med pravnimi akti, zlasti v primerih, ko avtomatizirane odločitve vplivajo na temeljne pravice posameznikov. Ena od osrednjih težav je zagotavljanje transparentnosti in razložljivosti odločitev UI sistemov ter zagotavljanje samodejne ocene skladnosti s politiko zasebnosti. V nalogi bomo preverili omejitve naravnih jezikovnih obdelav pri interpretaciji pravnih besedil. Velike jezikovne modele (LLM) in druge oblike generativne UI pogosto trenirajo na ogromnih in pogosto slabo dokumentiranih zbirkah podatkov, kar vodi do nejasnosti glede zakonitosti uporabljenih virov podatkov, vprašanj glede avtorskih pravic in morebitnih obdelav občutljivih podatkov brez pravne osnove. V primerjavi GDPR s kalifornijskim pravnim modelom CCPA in kitajskim zakonom o varstvu podatkov PIPL bomo ugotavljali možnosti za ustvarjanje čezmejnih regulatornih skladnosti UI sistemov, ter skladnosti z UNESCO priporočili o zagotavljanju etike sistemov UI in preprečevanju tveganj diskriminacije in marginalizacije posameznikov. Artifficial intelligence (AI) and machine learning systems are becoming a central part of modern digital services, with their capacity to process large volumes of data raising important questions concerning privacy protection and compliance with legislation such as the General Data Protection Regulation (GDPR). The purpose of this work is to analyze the key challenges in ensuring the compliance of AI systems with GDPR requirements and to evaluate the mechanisms, strategies, and legal frameworks enabled or mandated by existing European and international regulatory documents. AI, particularly in its generative and predictive forms, often relies on massive data collection and profiling, which raises dilemmas concerning the lawfulness of processing, the validity of consent, and the principle of data minimization. The EU has introduced a risk-based approach that emphasizes the reliability and transparency of AI systems; however, unresolved questions remain regarding the interplay between legal acts, especially in cases where automated decisions impact individuals’ fundamental rights. One of the core challenges is ensuring transparency and explainability of AI system decisions, as well as enabling automatic assessment of compliance with privacy policies. This thesis will examine the limitations of natural language processing (NLP) in interpreting legal texts. Large Language Models (LLMs) and other forms of generative AI are often trained on massive and poorly documented datasets, leading to uncertainties regarding the legality of data sources used, copyright issues, and the potential processing of sensitive data without a valid legal basis. By comparing the GDPR with the Californian legal model (CCPA) and China’s Personal Information Protection Law (PIPL), this work will explore the possibilities of establishing cross-border regulatory compliance for AI systems. It will also evaluate compliance with UNESCO recommendations on ensuring AI ethics and preventing risks of discrimination and marginalization of individuals. diplomske naloge AI GDPR compliance explainability privacy true true false V. Kores Slovenski jezik Angleški jezik Diplomsko delo/naloga 2025-12-22 18:58:15 2026-02-05 14:26:01 2026-02-06 03:21:53 0000-00-00 00:00:00 2026 0 Ljubljana 2026 Ljubljana X f., 68 str. 0000-00-00 Zaloznikova Objavljeno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 342.72/,73:004.8(043.2) 267495427 VS_Kores_Vasilij_2025.pdf VS_Kores_Vasilij_2025.pdf 1 DB6D0A8A7D91E033A674D30F5767717A ecb7c0de2eb4c760716e92e76732f70ff5fc9752c62452ad9b92a8e35088faa9 026e6c64-026e-11f1-b31a-00155d00019a https://dk.um.si/Dokument.php?lang=slv&id=203057 Fakulteta za varnostne vede 0 0 0