China will impose a “security inspection” on artificial intelligence tools, in full frenzy in the country of Internet giants to design ChatGPT-like tools.
The prowess of the American chatbot ChatGPT, launched in November and able to formulate detailed answers in a few seconds on a wide range of subjects, is being followed with interest in China.
The interface is not accessible in the country, but ChatGPT is the subject of countless articles and discussions on social networks, and local tech giants are vying to design equivalent tools in China.
The search engine Baidu was one of the first Chinese groups to position itself in this niche, joined by the champion of the Internet and video games Tencent or even Alibaba, a pioneer in e-commerce.
In full euphoria vis-à-vis so-called generative artificial intelligence, China wishes to regulate this technology.
Before being made available, products powered by generative artificial intelligence will have to “seek security inspection”, according to draft regulations published by China’s Cyberspace Administration on Tuesday.
The regulator, which submits its text for public comment before adoption, does not specify when the regulations will come into force.
“Socialist values”
AI-generated content must “reflect core socialist values and must not contain [d’éléments relatifs] to the subversion of state power,” the draft regulations read.
They will also have to respect the right to intellectual property, specifies the document.
The regulations are aimed at “the healthy development and standard application of generative artificial intelligence technology”.
This is “one of the strictest measures” to regulate ChatGPT-type artificial intelligence, Andy Chun, a technology expert at the University of Hong Kong, told AFP.
China aims to become the world leader in artificial intelligence by 2030, which will revolutionize a multitude of sectors including the automotive industry and medicine.
Baidu was the first in its country to announce that it is working on a local equivalent to ChatGPT.
Presented last month to the press, Ernie Bot, which works in Mandarin and is only aimed at the Chinese market, is currently only available in beta version.
E-commerce giant Alibaba unveiled its own on Tuesday, “ Tongyi Qianwen (a machine that knows everything).
” Be very carrefully “
In the race for artificial intelligence, the main difficulty in China for developers is to offer a high-performance conversational robot that does not deviate from the very strict framework permitted in terms of content.
So with this future regulation, companies will have to “be very careful” that each data used to feed their artificial intelligence “complies” with the directives, underlines Andy Chun.
In addition, “ensure the accuracy [des réponses] is difficult. No generative artificial intelligence system can do this today,” he notes.
China is at the forefront of regulating new technologies, which some see as a potential threat to the stability or power of the Communist Party.
After years of laxity, the authorities have since 2020 been intransigent with regard to the practices of powerful digital companies, in particular on issues related to personal data.
China already closely monitors its internet and media, with an army of censors erasing content that portrays state policy in a bad light or is likely to create unrest.
Social networks are subject to increased control.
China had already asked Internet giants last year to reveal their algorithms, a generally well-kept secret. Algorithms serve as the brains of many applications and services on the Internet.
The country also tightened in January the supervision of deep-faking (“ deepfake “), these digital manipulations of increasingly realistic images and which present a challenge in the fight against disinformation.