CONTU's doctrinal errors and their privacy implications: from copyright law to personal data protection in AI regulation

Abstract

This article critically analyses the doctrinal origins of the contemporary crisis in AI regulation. Examining the U.S. National Commission on New Technological Uses of Copyrighted Works (CONTU, 1974–1978), it argues that current attempts to apply copyright law to AI-generated outputs replicate CONTU's error — treating utilitarian objects as literary works. Particular attention is paid to the privacy dimension: mass training of generative models on LAION-5B and Common Crawl datasets violates the General Data Protection Regulation (GDPR) and renders data subject rights unexercisable. Drawing on contemporary scholarship, the article shows that the anthropocentric copyright model and GDPR instruments (text and data mining (TDM) exceptions, fair use, "legitimate interest") are structurally unsuitable for generative AI. The author proposes a sui generis framework separating human creativity from technical generation while integrating privacy by design, mandatory Data Protection Impact Assessment (DPIA), and AI content labelling.