Mindful not to stifle its budding AI scene, China will only require that the machine-generated content adheres to “core socialist values” in services available to the general public.
The as-of-yet provisional rules were jointly published by seven Chinese regulators, led by the Cyberspace Administration of China (CAC), the country’s all-powerful internet censor. They’ll come into effect on August 15th.
In addition to adhering to the state’s ideological line, generative AI service providers must ensure it does not produce subversive content or promote violence, ethnic discrimination, or pornography, among other things.
The regulations will also require providers to ensure that intellectual property rights are protected, advising developers to use legitimate data sources to train their models.
The CAC said that generative AI services would need a license to operate and take measures to stop generating any content deemed illegal before reporting it to the authorities and improving the algorithm.
The rules will apply to services provided to the general public and not in fields like research, as China seeks to encourage the development of a technology it sees as something it could rival the US with.
The regulations said they aimed to encourage “innovative” applications of generative AI and promote the development of related industries like semiconductors.
It will also offer a framework for Chinese companies that developed their own ChatGPT-style platforms, like Baidu’s Ernie Bot, to release their products to the general public.
Neither OpenAI’s ChatGPT nor Google’s Bard are available in China.