0100498910

open access

Version of Record

Current Developments on Labelling of AI-Generated Content: Comments on the AI Convention and the EU AI Act

English (英語)

Journal of International Cooperation Studies

33

59-67

神戸大学大学院国際協力研究科

2025-12

The note addresses the growing concern over disinformation fueled by AI-generated content（AIGC）, particularly images and videos, and makes a snap-shot analysis of the two legal instruments on AI: the Council of Europe’s AI Convention, an international treaty, and the AI Act, a regulation of the EU, both of which were adopted in 2024. As a background to this recent development, the note identifies a broad consensus among governments and international organisations that content provenance and authentication, understood as the identification of AIGC, are the key strategy to combat misinformation and disinformation of AIGC. There is also common acknowledgment that this is a matter of human rights, especially freedom of expression, according to the official discourse in international organisations. In this regard, the AI Convention promotes common principles rooted in human rights but leaves all the details to States Parties. It does not propose any labelling obligation for states or for AI businesses. In contrast, the EU AI Act establishes regulations regarding labelling of AIGC, and has specific rules regarding deep fakes, with clear obligations for AI providers and deployers. Other regulatory frameworks within the EU are likely to complement and support the AI Act, too. In the AI Act and indeed in any other instrument, while labeling AIGC enhances transparency, its effectiveness is reduced if different techniques are used on different platforms. The standardisation of these different techniques of labelling is needed. Its effectiveness also depends on the user AI literacy. There is also a need to heed the impact of labelling on privacy, and more generally, free speech.

紀要論文