To Label or (not) to Hide: the Transparency Principle and the Obligation to Disclose the Fact of Using Artificial Intelligence

The relevance of the study is underscored by deepfake proliferation, making the transformation of transparency into a legal obligation to disclose AI use particularly critical. The purpose of this article is to conduct a comprehensive study of the theoretical and legal foundations alongside global trends in implementing the duty to disclose the fact of using AI. The methodological framework comprises formal-legal and comparative-legal analysis, generalisation, and legal modelling. Analysing international instruments (EU, G7, OECD, UNESCO, Council of Europe, UN), the study highlights transparency as providing both functional information related to AI (internal, “invisible” technical processes) and outcome-related information (external manifestations “visible” to the public). Transparency rules across pioneering jurisdictions (Brazil, Canada, China, India, Japan, South Korea, US) are examined, with particular attention to Article 50 of the EU AI Act. The main challenges of legal regulation regarding disclosure are identified: the diversity of AI applications, the involvement of a wide range of actors beyond AI providers, the lack of standardized labeling and notification formats, and the underdevelopment of legal liability mechanisms in the field of information technology. The paper justifies labelling AI-generated content (including deepfakes) and identifying all AI interactions (e.g., conversational chatbots) through technical markings (visible and invisible), verification tools, and public notification methods. Proposals are formulated to enshrine transparency obligations in Ukraine’s future AI legislation. The findings hold theoretical and practical value for software developers, providers and deployers of AI systems, lawmakers aligning Ukrainian legislation with EU standards, and internet users seeking enhanced digital media literacy.