Shanghai Intellectual Property Court

Headlines · 1

1. Oct 15
   Explainer: LLM training-data compliance risks in copyright, personal data and trade secrets

   36Kr republished an article by Sui Jingxuan, originally from the WeChat account Zhichanli, that maps large-model training-data risk across three dimensions — third-party copyright, personal information, and patents and trade secrets — and proposes a framework of classifying data by type and controlling it across the full data lifecycle. The piece cites 2025–2026 cases including US authors collectively suing six AI vendors, a Munich court finding OpenAI infringed by training on lyrics without authorization, and the Shanghai Intellectual Property Court concluding China's first AI large-model copyright infringement case. The author argues that China's Article 24 of the Copyright Law uses a closed list of fair-use exceptions that model training does not readily fit, and cites industry views that the practice falls under neither fair use nor an expanded statutory licence, leaving its legal status highly uncertain. Recommended paths include licensing or prioritizing open-source and public-domain data, anonymizing personal information as the first option, and screening data sources for patent and trade-secret content.

   36氪 人工智能 · 🔥 0

Experience and discussion from the community

Share my Shanghai Intellectual Property Court experienceAsk about Shanghai Intellectual Property Court