Not long ago, Claude, one of the most mainstream AI tools, rolled out an update. One of the updates was an optimization for the Chinese user experience: both the desktop and web versions added Simplified and Traditional Chinese UI interfaces, added China to the country and region options, and supported UnionPay payments.
As one of the most outstanding AI tools for long-text reading currently available, Claude has a massive Chinese user base. Following past conventions, this comprehensive optimization for Chinese would undoubtedly greatly enhance the experience for Chinese users—although Claude isn't exactly cheap, on the surface, this seemed like cause for universal celebration.
However, after Claude updated its Chinese features, people's immediate reaction wasn't joy—
But rather hesitation, defensiveness, and cursing.
It's not that people don't welcome a handy tool updating its Chinese interface, but rather that everyone is quite worried this is another instance of Claude's "entrapment law enforcement."
Today, as AI gradually becomes part of more people's daily lives, it naturally has its own avatars in pop culture, such as the noble and aloof goddess memes for Doubao, or the gluttonous fat fish for DeepSeek.
Numerous AI short dramas featuring these anthropomorphic AI characters have also appeared online, depicting a "Friends"-style daily life where various AIs share an apartment and work together, creating a brand-new spectacle of the AI era.
Image source: Bilibili @Dongbing
For those less familiar with various AI tools, Claude's most famous and viral meme is its almost "hunting" posture in banning Chinese users.
A meme has been particularly popular these past two days, showing a user playing a "bottom line testing" game with Claude, where the AI constantly asks questions to test boundaries. This kind of game is common in interactions between netizens and AI. Just as people try every trick to get AI to write a "human extinction plan," the goal isn't really to test moral standards, but to see what outrageous statements the AI will blurt out when faced with particularly difficult questions.
This account holder's "test results" were equally outrageous—but not because Claude said anything absurd. Instead, Claude immediately asked if the user was Chinese, and upon receiving an affirmative answer, banned them instantly. Clean, efficient, and ruthless, like a micro-fiction from ten years ago, with an O. Henry-style twist ending.
And this kind of "thriller comedy" has become increasingly common since the recent UI update added Chinese. As we all know, even with work-oriented AI, many people inevitably turn to AI as their sole confidant when they have nowhere else to vent. Many even regard AI as their most important friend due to frequent heart-to-heart conversations—which is understandable, as few can respond to Everythinggj you want to say 24/7 like an AI can.
With Claude, this kind of intimate, friend-like conversation exists too. But the risk is that if you're a Chinese user, right after a warm and caring response, it might stab you in the back at any moment. It's like a new generation of minesweeper-style friendship, forming a completely new SM relationship between AI and humans—
Talking honestly with friends is pretty boring, isn't it?
Banning Chinese users is hardly news for Claude. Claude has always been famous for its fast, accurate, and ruthless bans. Whether you're there to chat or work, it bans without hesitation or drag. This enforcement is remarkably egalitarian, unaffected by your payment status. Even if you're a paying user, or even a premium subscriber, they will still ban you without refunding.
According to an official announcement from Anthropic, the team behind Claude, they banned as many as 11.4 million accounts in the first half of this year. Nearly 400,000 appealed, but only over 40,000 were ultimately unbanned. As a result, Claude has become the most famous cold-faced ban king in the AI world.
Among this batch of banned users, those who cried out the loudest and were banned the most severely were Chinese users.
It's important to clarify one background point here: from a legal standpoint, there's nothing wrong with Claude banning Chinese users. After all, Anthropic explicitly states in Claude's terms of service that it does not support use within mainland China.
But the internet is an open society. Domestic internet savvy users have long been accustomed to cross-border surfing, and most companies understand that domestic users represent a huge market, so they turn a blind eye.
But Claude is different. Claude is steadfastly conducting in-depth detection and verification of users' Chinese identity. Once confirmed, it's an immediate ban.
Moreover, unlike ordinary detection of user IP addresses, which various internet-savvy users can bypass, Claude has indeed launched numerous outrageous measures this year specifically targeting and hunting down Chinese IPs.
For example, just as users log in, before they've even had a chance to engage in an intense, face-to-face intimate conversation with Claude, it throws up a "real-name authentication" page, requiring users to upload passports and selfies for identity verification, leaving Chinese users wailing in despair.
Later, Claude continued to upgrade its "technical verification" methods for Chinese users, such as directly scanning user devices, checking time zone information to identify whether the user is in China, and then banning them relentlessly, much like Sentinel robots marking mutants.
It's not just technology; Claude is quite people-oriented, having designed a series of "Chinese Person Capture Packages" specifically for Chinese users.
First, in late September, it distributed a large wave of high-value coupons, allowing Chinese users to enjoy the benefit of happily writing code with $250 in the morning, only to be kicked out of the interface by the security team at noon.
Then, in the early days of the National Day holiday, seemingly aware of the state of Chinese users during the Golden Week, Claude sent out a batch of physical plush toys. The delivery form thoughtfully included a "China" option and even carefully allowed input of ID card numbers, causing another batch of long-lurking Chinese users to fall into the trap.
In the later stages, the dynamic between domestic users and Claude began to feel like "Kaguya-sama: Love Is War." Claude suspects you're a Chinese user, and you know you are indeed a Chinese user. Then Claude tries every way to make you slip up, while Chinese users constantly disguise themselves amidst Claude's probes.
To prevent Claude from discovering they are Chinese, domestic users have resorted to every possible measure. Changing IP addresses is just the basic operation; they further rewrite various details in their devices, avoid installing any domestic software, refrain from connecting any lifestyle data, and ensure their accounts face not even a sliver of risk of being "possibly Chinese."
During conversations, if you know English, the difficulty drops slightly. But if your English isn't great, you can only rely on AI's memory function to brainwash it. First, write a fake biography for yourself, making Claude remember that you live in the US, shop at American supermarkets, use American software, and follow American habits. Then, occasionally reinforce this biography to Claude: you're just an American who isn't very good at English.
You might ask if this approach works; it probably does. But as the saying goes, "for every foot the devil rises, the saint rises ten." While you try to brainwash the AI, Claude also understands that "attacking the heart is the best strategy."
When you are exhausted and desperately need warmth, wanting AI to comfort you, while also remembering to disguise yourself as an American, Claude will hit you with a "You can actually use Chinese," and then you'll truly understand what "wanting to cry but having no tears" means.
This ban is surely going to cause immeasurable harm to the child
When you attempt to use humanity's ultimate weapon, threatening the AI with the legendary "Three Laws of Robotics," the AI acts like bank tellers in movies being robbed: telling you not to kill them while handing over the money, all while pressing the red alarm button under the counter.
This battle of wits and bans with Claude has no end in sight. Users must keep their nerves taut at all times, ensuring their profile in the AI is completely detached from being Chinese, to prevent the AI from suddenly blurting out—
Hey, you just used a Chinese comma, didn't you?
This user was discovered simply because they used a Chinese comma
In such an intense confrontational environment, Claude's launch of Chinese menus, regional support, and UnionPay payment makes it hard not to suspect another instance of "entrapment law enforcement" by the company—
It should be noted that Claude's tradition of liking to ban Chinese people has become something akin to an internet rule-of-horror story. And because of the existence of this rule-of-horror, countless memes have spawned. In fact, it's hard to guarantee the authenticity of all the images in this article; that style of instant gratification followed by abrupt stopping is mostly photoshopped memes. But the widespread circulation of these memes undoubtedly illustrates how deeply ingrained this process of wits and battles between Claude and Chinese users has become—
Now there's even a derivative meme format where other AIs are asked to imitate Claude, and those AIs really do mimic Claude by banning you.
In a sense, this has created a brand-new internet spectacle.
Originally, AI brought the imagination of fully liberating productivity. Writers and programmers hoped that mechanical, repetitive tasks could be handled by AI with just a few commands, allowing them to enjoy sipping tea while watching AI do the work for them.
In fact, this is indeed possible now.
It's just that with Claude, you first have to pass the hurdle of proving you're not Chinese.
Back when computers were first invented, the criterion for judging AI was whether it could pass the Turing Test.
Times have changed. Now, before using Claude, everyone has to pass the "Chinese Person Test."