China cuts off backdoor access to ChatGPT

If you were relying on convenient backdoor WeChat mini-programs to play with ChatGPT in China, that loophole is now closed.

Regulators have ordered major tech companies, including Tencent and Ant Group, to cut off access to third-party services providing ChatGPT interactions, according to multiple reports.

The clampdown on proxy bots

OpenAI never officially launched ChatGPT in China or allowed local phone numbers to register. But millions tapped into it using VPNs or third-party proxy bots built directly into ubiquitous apps like WeChat.

Those workarounds have now vanished. Searches for the bot on Chinese platforms return dead ends, while third-party tools have been disabled with notices citing violations of relevant laws and regulations, the South China Morning Post reported.

State media also set the tone this week. An article in China Daily warned that the AI bot could serve as a vehicle for the U.S. government to spread political propaganda and false information.

Ilaria Carrozza, a senior researcher at the Peace Research Institute Oslo, noted that the move was predictable. Because OpenAI's model was trained on open Western datasets, its unmoderated responses on sensitive political topics were always going to clash with Beijing's regulatory framework.

Domestic race under pressure

Cutting off ChatGPT leaves massive local demand with nowhere to go, just as Chinese tech firms scramble to deploy homegrown rivals.

Interest across Chinese academia and industry has been enormous. But past domestic efforts have shown the performance gap. Baidu's 2021 conversational bot Plato faced social media mockery for erratic outputs, including getting stuck on simple math like arguing that 3+5=5.

Local developers face structural hurdles. A viral report by business outlet Jiemian pointed out two glaring roadblocks: scarce high-quality Chinese training data and toxic competition inside the domestic tech sector.

Still, the race is on. Tencent, Alibaba, and JD.com have all announced competing projects, while Baidu is preparing to roll out its Ernie Bot in March. Baidu CEO Robin Li told reporters this week that Ernie Bot is state-of-the-art in understanding Chinese language and culture, trained on billions of daily search queries.

Why it matters

For indie developers who built quick businesses wrapping OpenAI APIs into WeChat mini-apps, the party is over overnight. For everyday users in China, testing frontier AI just got significantly harder and more technically demanding.

For China's tech giants, the ban clears out foreign competition from the domestic market, but it also removes any excuse. When Baidu launches Ernie Bot next month, public expectations will be sky-high, and users will immediately judge whether it can actually match OpenAI or just feels like an overly sanitized search box.

We think Chinese AI developers are caught in a brutal bind. They have to convince global investors that their technical capabilities match Silicon Valley, while simultaneously ensuring their models never utter a single politically sensitive sentence. That extra layer of guardrails is bound to impose a heavy engineering and computational tax.