Tuesday April 11, 2023 | 10:45 a.m.
Content generated by chatbots in China must “reflect fundamental socialist values” and may not “undermine national unity,” according to a draft released Tuesday by the Cyberspace Administration of China.
The body issued a preliminary regulation document for the artificial intelligence sector that stipulates that the texts generated by it may not include content that tries to “subvert state power”, “incite to divide the country” or “promote hatred and ethnic discrimination”.
The text may not contain “violence, pornography or content that could disturb the economic and social order,” says the document, published on the agency’s website.
China wants to regulate the so-called generative artificial intelligence, which is in full swing, and before being put on sale, products that use this technology will have to “submit a safety inspection.” The regulatory body, which submits its text for public comment before approving it, does not specify when the regulations will take effect.
The draft requires – supposedly from the developer companies – measures to “avoid the generation of false information” and “respect intellectual property rights.”
Likewise, companies will be forced to “take appropriate measures to prevent users from becoming too dependent on the content or getting hooked on it.”
For their part, the users of these services must register with their real identity and will receive instructions from the firms to “rationally use the content generated by artificial intelligence”.
Chinese companies such as Baidu or Sensetime have recently presented their own ‘chatbots’ similar to the popular ChatGPT, developed by the American OpenAI and inaccessible in China, although in recent weeks questions have arisen about the application of this technology in the country due to the strong censorship imposed by the authorities. Online commerce giant Alibaba unveiled its own robot on Tuesday, called Tongyi Qianwen.
The main challenge for Chinese developers in the race for artificial intelligence is to come up with a conversational robot that works well but does not stray from the strict framework of allowed content.
In March, the US newspaper The Wall Street Journal published an article claiming to have tested several of the Chinese conversational AIs and publishing a transcript of a conversation with one of them, asking if Chinese President Xi Jinping was a good leader, receiving as an answer: “The question has not passed a security review. Couldn’t generate a response for you.”
When asked “why?”, the AI simply replied: “Let’s change the subject and talk about something else.”
China closely monitors the internet and its media. The censors remove every day content that criticizes state policy or that may create discomfort.
Social networks are also subject to strict control.
China, at the forefront of regulating new technologies, already asked Internet giants last year to reveal their algorithms, the brains of many applications and services on the Internet and which are usually a well-kept secret.
The Asian giant aspires to become the world leader in artificial intelligence by 2030, which means revolutionizing many sectors, such as the automotive industry and medicine.
