The Legal Framework for Source-Based Governance of AIGC Disinformation: A Comparative Perspective Between China and the EU

Abstract

The rapid iteration of generative AI has engendered complex forms of disinformation of AI-generated content (AIGC Disinformation), which constitutes a systemic risk spanning technical, legal, and social dimensions. Comparative research of Chinese and EU legal frameworks indicates that effective source based governance of AIGC Disinformation requires further development and harmonization. In terms of training data governance, China emphasizes the legality of data sources, quality assurance, and security management to ensure the authenticity of outputs. The EU focuses on transparency and traceability mechanisms, alongside data quality management systems, to mitigate AIGC Disinformation risks. Regarding AI model risk management, China has established lifecycle-based security assessment mechanisms aligned with national recommended standards, applied on a voluntary basis. The EU employs a risk based regulatory framework, imposing stringent obligations particularly on general-purpose AI (GPAI) models with systemic risks and models used in high-risk AI systems. Despite differences, both jurisdictions treat training data governance as the first line of defense and AI model risk management as the second line of defense. Both also emphasize clarifying the primary responsibilities and liabilities of AI providers. Looking forward, China and the EU should promote global consensus on AIGC Disinformation governance through collaborative frameworks and mutual technical recognition, fostering the high-quality development of generative AI.