Abstract:
This paper examines how Chinese copyright law can reconcile authors’ rights with generative-AI innovation through a dual phase duty-of-care regime. It proposes recognising AI training as a form of fair use or a copyright exception conditioned on “copyright alignment” during training – requiring lawful data sourcing, dataset governance, and anti-memorisation safeguards – and establishing proportionate duties of care during deployment, including filtering, watermarking, and “notice and necessary measures” mechanisms under Arts 1195 and 1197 of the Civil Code of the People’s Republic of China. “Necessary measures” encompass deleting infringing content, blocking prompts, and optimising models to prevent recurrence. Drawing on Chinese cases and comparative regimes, this paper argues that technically grounded obligations can mitigate infringement risks while fostering innovation and responsible AI governance in China.