Law / China

Interim Measures for the Management of Generative AI Services, Articles 7 and 8 (training data)

Interim Measures for the Management of Generative AI Services (生成式人工智能服务管理暂行办法), 2023, Arts. 7-8

In force since .

An AI training data rule.

As of .

What it requires

  • Use data and base models from lawful sources when you pre-train or fine-tune a generative AI service offered to the public in China.
  • Do not infringe others' intellectual property rights in the training data you use.
  • Obtain the individual's consent, or meet another circumstance that a law or administrative regulation sets, before using personal information in training data.
  • Take effective measures to raise the quality of your training data and to strengthen its truthfulness, accuracy, objectivity and diversity.
  • When you label data during development, set clear, specific and operable labeling rules, assess labeling quality, check a sample of the labeled content for accuracy and give labelers the necessary training.

These lines are LexLint's reading of the law's text, as of .

What this law does

Drafted with AI

Drafted with AI from the cited sources under the direction of UnGovr staff. UnGovr holds editorial responsibility for this page. How this site is made

Research summary

Legal information, not legal advice. This is LexLint's own research summary of a public legal source, and it creates no attorney-client relationship. For decisions that matter, consult qualified counsel in the relevant jurisdiction. About LexLint

Article 7 requires a provider to carry out pre-training and optimization-training data processing in accordance with law, using data and base models from lawful sources. Article 7, item 2 provides that where intellectual property is involved, a provider must not infringe the intellectual property rights others hold under law.

Article 7, item 3 requires a provider, where personal information is involved, to obtain the individual's consent or to meet another circumstance that a law or administrative regulation sets. Article 7, item 4 requires a provider to take effective measures to raise the quality of training data and to strengthen its truthfulness, accuracy, objectivity and diversity.

Article 8 requires a provider that labels data during generative AI research and development to set clear, specific and operable labeling rules, assess labeling quality, check a sample of labeled content for accuracy and give labelers the necessary training.

When LexLint raises it

When your app profile says your app trains models.

Back to the example  ·  Lint your app