China Proposes New Rules for Firms Offering AI Services Like ChatGPT, Including Banned Training Data

China published proposed security requirements for firms offering generative AI services like ChatGPT. This includes a blacklist of banned training data.

Banned training data includes information advocating terrorism, violence, overthrowing China's system, damaging China's image, and undermining unity.

Censored internet content in China also cannot be used to train models.

Firms must get consent from people whose personal data is used for training.

Guidelines aim to avoid IP violations. China wants to be a world AI leader by 2030.