China intends to test chatbots before their release

The China Cybersecurity Authority wants to test the security of generative artificial intelligence systems before they are released by developers. Regulator announced on the start of public hearings on the need to introduce such a measure.

According to the report of the department, according to the authorities, products using generative artificial intelligence (AI) must comply with the current requirements of Chinese law. Separately, it is emphasized that the content created by such AI should not contradict the norms of public morality, and also “contain calls for undermining state power, overthrowing the socialist system, inciting the division of the country, undermining national unity.” The bill also suggests that developers will be fully responsible for the content that their AI creates, as well as for the mishandling of personal data that AI collects during the content creation process.

At the same time, it remains unclear how the regulator will deal with those systems of generative artificial intelligence that have already been released by companies and, accordingly, have avoided government checks. Yes, Baidu introduced of his ERNIE chatbot back in mid-March. On Monday, April 10, there was a presentation of the SensaNova neural network from SenseTime. And today, April 11, Alibaba Cloud introduced its product – Tongyi Qianwenwhich supports Chinese and English and will initially be built into DingTalk corporate messenger and Tmall Genie smart home devices.

Kirill Sarkhanyants