BEIJING – AI Creative Chat, Chat Intelligence AI, and Yiqi AI Chat.

These names might seem unfamiliar to most, but all three claim to be the Chinese version of artificial intelligence (AI) service, ChatGPT.

Accessible through super app WeChat, these ChatGPT proxy services are among a plethora of alternatives available on the Chinese Internet, to bypass a government block on the service. These include proxy services, the sale of ChatGPT accounts and even a query service.

Microsoft-backed OpenAIs ChatGPT services are not available to Chinese users. Access to the OpenAI website is still available, but attempts to open the chat function on a Chinese Internet connection results in an error code.

With the rise of artificial intelligence, Beijing is now faced with a dilemma between making such services widely available, and its need to control access to information.

Existing behind what is known as the Great Firewall, Chinas Internet is in a parallel domain from the rest of the world. It is tightly controlled, with armies of censors hired by tech firms to ensure information deemed politically sensitive is scrubbed. Users who violate such rules can be banned or even criminally prosecuted.

Services like Google, Facebook and Twitter are blocked in China, along with the websites of several news outlets including The Straits Times.

But as it promotes AI development, Beijing has to contend with what it says are issues to do with controllability.

With a large language learning model like ChatGPT, whose responses are trained based on information available across the global Internet, Chinese users can easily access data deemed sensitive by the government.

Promoting AI development while trying to strengthen ethical norms is something the government is looking at, said Minister of Science and Technology Wang Zhigang on Feb 24.

In our country, we have taken corresponding measures in terms of ethics for any new technology, including AI technology, to ensure that the development of science and technology is beneficial and poses no harm and to leverage its benefits better, he said.

Shortly after that, regulators blocked access to ChatGPT and proxy services.

But that has done little to deter Chinese users.

After borrowing my friends (ChatGPT) account, Im hooked to how easy it has been for research, especially writing up drafts of papers that I can then translate, said undergraduate Rose Zhu, who had tried the service before it was blocked. More On This Topic Chinas Baidu scraps public launch for ChatGPT-like product China's students leap over 'Great Firewall' to get homework help from ChatGPT The 20-year-old history major student said this led her to pay 29 yuan (S$5.60) for a one-month subscription to Yiqis proxy service to continue using the service when doing research. Its unpaid service limits the number of queries per day.

It is definitely slower, and its answers can occasionally be a bit fuzzy, but I suppose thats the compromise since I dont have to go through a VPN, she said.

Other friends who bought actual ChatGPT logins from online merchants have had their accounts suspended after OpenAI started clamping down on account sharing, she said, adding that a potential sudden crackdown was the reason why she bought only a months subscription.

I know its a grey area, but we need to get used to using such products if not we will be left behind, she said.

On e-commerce platforms like Taobao and Xianyu, account logins to ChatGPT and subscriptions to proxies were available for sale but have since been taken off after the regulators told tech firms not to allow access to such services on their platforms.

But enforcement is uneven: Over the past week, The Straits Times was able to register for three proxy services through WeChat. They function as vehicles feeding queries to ChatGPT, then delivering responses back to the user.

Response times are slower. For instance, a request to write a 500-word essay on Asean-China relations took less than 20 seconds on ChatGPT via VPN, but the same request took about half a minute on Yiqi AI. The same question crashed the Chat Intelligence system.

They apparently even self-censor.

A query to Yiqi AI about Chinese President Xi Jinping resulted in an error message because the query contained sensitive terms. There were no such issues when it came to American political leaders.

Some also appeared to be using limited access versions and would present errors when asked more technical questions. For instance, when asked to check the source code for a program to collect the top 10 headlines from a news website, one proxy said they could not answer the question and the other hanged. More On This Topic Why China didnt invent ChatGPT ChatGPT frenzy sweeps China as firms scramble for home-grown options By mid-week, following yet another round of enforcement, AI Creative Chat was suspended by WeChat for offering services not yet allowed under (WeChats) user terms agreement.

Queries to all three service providers went unanswered.

Meanwhile, on social media platforms like Xiaohongshu, some are offering asking services where users send questions through an intermediary who inputs the questions into ChatGPT on their behalf, for up to 50 yuan a question, depending on complexity.

One such provider, who spoke to ST through a secure messaging app, said he is based overseas and was simply trying to find a way to monetise his access.

Its a legal grey area and from what we see, the authorities are not coming down too hard for now, but we all know when the government decides to crack down, it will be fast and brutal, said the service provider, who wanted to be known only by his last name, Zheng.

So far, about 10 people have used his service, mostly out of curiosity, with one asking for help with a coding question, he said.

Mr Zheng declined to say where he is based and how he got access to a ChatGPT account, but his user profile on Xiaohongshu showed his IP address to be based in Sichuan province.

Beyond the control of information, concerns about data security is another reason for China restricting access to the service, said Dr Shaoshan Liu, a US-based tech policy expert and founder of PerceptIn, a firm specialising in autonomous driving and robotics. More On This Topic China prepares to police AI as ChatGPT frenzy spreads Baidus ChatGPT-like app will revolutionise its search engine, says CEO In recent years, China has been concerned that data about how its 1.4 billion people live could get into the wrong hands, resulting in national security threats. This manifested itself in 2021 when a ride hailing firm was penalised for violating data security rules. It resulted in sweeping new laws governing how tech firms handle user data, but there are still worries that engineers checking source code could accidentally reveal confidential information.

Other countries are also restricting access. Italy has recently said it will temporarily restrict access to ChatGPT because of privacy concerns and numerous other European Union nations including Spain and France are also considering similar regulations.

It is likely that in the coming months, there will be more regulations on large language model learning products globally, and China will be no exception, said Dr Yu Jianing, president of Uweb, a firm specialising in Web3 technology now based in Singapore.

This comes as Chinese tech firms star to roll out their own AI-generated content products.

Firms ranging from Baidu, Alibaba, Tencent and iFlyTech all say they are developing their answer to OpenAIs service but nothing close to ChatGPT has emerged so far.

Baidus Ernie Bot was launched to lacklustre response in March when instead of a live demonstration as widely hyped, the service was launched via a pre-recorded video and developers have rescinded public access.

Dr Yu noted that excessive rule-making, while necessary in offering guardrails for benign interaction between innovation and governance could also stifle growth.

Therefore, it is necessary to balance the relationship between innovation and regulation, and formulate appropriate policies and regulations to manage the development and use of AI products, he said. More On This Topic What is ChatGPT-4 and how to use it now: Everything you need to know Chinese state media, AI companies warn of risks in ChatGPT stock frenzy