China has decided to tighten its grip on ChatGPT. Although OpenAI’s generative AI was not officially available in the Asian giant, in practice there were those who had managed to access popular bots through VPN —virtual private networks— and some external developers had even created programs with the service. Now the authorities of the country have wanted to go one step further and cut off the access: they have ordered their main technological companies to eliminate any solution that is free from the service.
El motivo: quiere evitar “erroneous information” made in USA.
What has China decided? Reinforce the ban on ChatGPT among its citizens. As detailed by Nikkei Asia, China’s regulatory authorities have ordered large technology companies not to offer ChatGPT services to the public. In fact, Beijing has given instructions to Tecent Holding and Ant Group, a subsidiary of Alibaba Group Holding, so that they do not offer access to bots on their platforms, both directly and through third parties.
But… Does ChatGPT work in China? No. And yes. The chatbot, created by OpenAI and supported by Microsoft, both based in the United States, is not officially available in China, but as it is the norm, there are those who have managed to access it. The Guardian there are two ways: private networks and programs from external developers that have now disappeared from the accounts of WeChat, the popular Chinese instant messaging app launched by Tencent.
What is the measure supposed to do? From a Hong Kong newspaper South China Morning Post assures that since November, “dozens of public accounts” had appeared in WeChat that incorporated ChatGPT through an application programming interface (API). Thanks to them, Chinese users were able to bypass the veto and use the bot without a foreign phone number or VPN.
On Wednesday, the Hong Kong media reported that at least some had been deactivated, including Yibai Technology, Shenlan BL, AI Duihua or ChatGPT Robot. In this case, the concrete account that has been suspended for “violating the relevant laws and regulations” is the last one. Other apps signal that their services have been disconnected due to “relevant commercial changes” or political changes, without providing more details about the reason. DingTalk, for example, Alibaba’s business collaboration platform, would also have eliminated a guide published at the beginning of me with instructions on how to add a ChatGPT bot to group chats.
And what is the reason? International media point to a motive a priori as simple as peliagudo: to control the flow of information. ChatGPT is finally a tool created for a US company, a model – comments a The Guardian Dr. Ilaria Carrozza, from the Peace Research Institute of Oslo—”training in open information based in western countries”.
“Potentially plant muchos problems [para Pekín] because people could have used it to raise questions about sensitive issues, such as the abuse of human rights in Xinjiang, Taiwan, the Diayou Islands”, he underlines. A few days ago, Chinese state media was slipping that the US would be using the chatbot to “spread disinformation and manipulate the public opinion” and specifically indicated their answers about Xinjiang. When asked for information about this region of northwestern China, ChatGPT describes the alleged abuses suffered by ethnic minorities, including massive detentions and re-education camps.
What is the context? Hace una semana China Daily, a newspaper linked to the Chinese Communist Party, published an extensive opinion article in which it warned of the networks that planted artificial intelligence and complained, already from the same title: “The rise of ChatGPT calls for a regulation of the AI”. One of the risks that its author, Calving Tang, pointed out was precisely “the dissemination of false information” and “fraudulent contents that appear to come from a reliable source”.
The measure of the Chinese regulators also coincides with the race of some of its companies and universities in the country to develop AI chatbots. One of the most advanced would be from Baidu’s Ernie Bot program, which already launched Plato in 2021.
Cover image: Owen Winkel (Unsplash)
