A Brief History of Claude Account Bans: Anthropic's Three Years of Bans and the Cat-and-Mouse Game with Chinese Users

By Letter List

In Anthropic's publicly disclosed IPO prospectus, a special source of revenue was revealed.

In the first nine months of 2026, about 5% to 7% of revenue came from customers later determined to have violated usage policies or commercial terms.

Anthropic's revenue in the first three quarters was about $33 billion. By that calculation, these customers contributed roughly $1.65 billion to $2.31 billion in revenue.

Chinese netizens clearly contributed a considerable share of this "banned account" revenue.

Not long ago, another netizen's Claude account was banned, and the ban was somewhat baffling.

This netizen had been using the account for more than a year, and because he knew that mainland China was not within Claude's service area, he had always been prepared. At first, he even only dared to chat with Claude in English; later, after using it for a long time, he gradually grew bolder and began using Chinese with confidence.

But he never expected that the reason Claude ultimately gave for the ban was that the system believed he was under 18.

The netizen immediately found it a bit hard to take.

"I can accept it if it banned me because it thinks I'm Chinese; in a sense, that would count as dying of old age. But what does it mean to ban me because you think I'm underage? That counts as sudden death, okay!!"

Interestingly, Claude has recently rolled out a series of product changes aimed at Chinese-speaking users.

Claude not only added support for Simplified Chinese and Traditional Chinese interfaces, but UnionPay also appeared as an option on the payment page, and the billing region could even be selected as China.

Some people think that Anthropic has finally remembered to collect money from Chinese users.

Others joked: Could this be a plan to "lure domestic users in and then kill them"?

01

Claude was never open to mainland Chinese users from the very beginning.

On March 14, 2023, Anthropic officially launched Claude. This AI assistant was initially mainly open to partners, and ordinary users could not yet register freely. On July 11, Claude 2 was released, and Anthropic finally launched the public-facing Claude.ai website, with only the United States and the United Kingdom opened first.

After that, Claude gradually entered more countries and regions, but mainland China was never included.

For mainland netizens who wanted to use Claude, the first problem was how to get access to it at all.

From overseas network routes and account registration to payment methods, domestic users gradually figured out a set of methods. But successful registration was only the first step; the Claude they had worked so hard to use could be banned at any time.

As early as August 2024, someone on LINUX DO posted "Claude Account Ban Appeal Template Form," specifically teaching netizens how to write English appeal letters to Anthropic.

This guide was written in considerable detail. From the email subject and grounds for appeal to how to explain abnormal login activity on the account, everything was covered.

The template even deliberately designed a set of talking points: that the user was an American user who had to use a proxy network to access Claude only because work required a business trip to China.

In the comments, some complained that their appeal emails went unanswered, while others shared successful experiences.

This account had not been used for eight months. After hearing that Claude had recently been banning accounts severely, he went to take a look and found that his own account had also been banned. So, referring to the appeal template shared in the community, he wrote a short essay to Anthropic, and unexpectedly, it actually worked.

He also posted the restored chat history, whose records stopped eight months earlier.

However, although the account was recovered, he still felt uneasy and could not help sighing that now he only dared to use it through a relay station.

By 2025, the tug-of-war between Chinese users and Claude over account bans had produced all kinds of stories that left people not knowing whether to laugh or cry.

In January 2025, LINUX DO user NoobNoob experienced a baffling account ban.

He had already used Claude for more than a year and was a long-term paying Pro user. After his old account was banned, he wrote an appeal email to Anthropic while registering a new account.

The new account used the same network route and payment card as before and successfully activated Pro membership.

Just as he was preparing to give up on Claude and switch to Gemini, the appeal for his old account suddenly succeeded.

Anthropic sent an apology notice, and he resubscribed using the same network route and payment card. This time, everything was actually normal.

After all that back and forth, he himself could not figure it out and could only sigh: "Very confusing, I don't understand what's going on with Claude either."

There are many similar experiences.

Some people's Claude accounts were banned after half a year of use, and newly registered accounts did not even last a month; some had not logged in to Claude for more than half a year and suddenly received a ban notice; others used the same computer and the same network route, yet several accounts met completely different fates.

Appeals were also a matter of luck. Some waited months without a reply, while others successfully restored their accounts by writing only one sentence: "I use it for coding."

Ordinary users could not figure out Claude's ban rules, while Anthropic kept adding new thresholds.

In April 2026, Anthropic began "checking household registration" for some Claude accounts, requiring users to submit government-issued documents such as passports and driver's licenses to complete identity verification.

As a result, after the new threshold appeared, the account business surrounding Claude also upgraded accordingly.

On June 26, the American magazine Wired published an investigation specifically examining how Chinese users bypass Anthropic's regional restrictions. The report found that from Claude accounts on Taobao and Xianyu, to account trading in Telegram, to API relay stations that specifically forward Claude requests for domestic users, a market has already formed around Claude's access restrictions.

After identity verification was introduced, ready-made accounts claiming to have already passed identity verification even appeared on Telegram.

Faced with these endless methods, Anthropic also confirmed to Wired that the company is using constantly changing detection systems, including identity verification, and is specifically identifying and cracking down on proxy networks that provide Claude access services.

The official side is busy strengthening restrictions, while Chinese netizens continue to study how to keep their accounts.

A diverse set of "account preservation superstitions" gradually spread.

Some studied IP addresses and payment cards, some worried that their computer time zone, system language, and browser fonts exposed their location, and others summarized a pattern that using Claude at 5 a.m. made bans more likely.

Recently, even the technical reports published by Anthropic itself have become objects of study for netizens.

On October 13, Beijing time, Anthropic published an anti-distillation technical article detailing how the company identifies and prevents other organizations from extracting Claude's capabilities at scale.

The official side revealed that the new system uses a two-stage classifier, with a false positive rate below 0.1% for detecting requests for specific distillation behavior; for accounts that continuously attempt to extract model reasoning content, Anthropic may limit their usage quota or even directly ban them.

The article also mentioned that Anthropic has already banned tens of thousands of accounts related to distillation, and some accounts were banned less than two hours after starting related activities.

As a result, Chinese netizens immediately began reverse research.

Less than four hours after the article was published, AI blogger Guizang posted: Anthropic disclosed how it fights distillation, but also, to a certain extent, told ordinary users how to avoid being banned.

In 2023, Chinese netizens were still studying how to use Claude; in 2024, they began teaching each other appeal essays; and by 2026, even the anti-distillation technical report published by Anthropic itself could be read by netizens as a reference for avoiding bans.

Three years later, the threshold for using Claude has become higher and higher, and domestic users have more and more countermeasures.

This cat-and-mouse game around Claude bans is clearly far from over.

02

When it comes to why Claude has never opened to mainland China, one has to mention Anthropic's CEO Amodei.

Interestingly, he worked at Baidu more than ten years ago.

In 2014, Amodei joined Baidu's Silicon Valley AI Lab and studied deep learning and speech recognition under Andrew Ng.

At the time, he also participated in developing Baidu's Deep Speech 2 speech recognition system and became the first author of the related paper. This system could recognize English and Mandarin and was a fairly representative end-to-end deep learning speech recognition achievement at the time.

More than ten years ago, Amodei was still at Baidu researching how to make AI understand Mandarin; now, the Anthropic he leads still does not open Claude to mainland Chinese users.

This inevitably makes people curious: what exactly did he experience at Baidu back then?

Coincidentally, in June this year, a Bloomberg host really did ask Amodei face to face: did your experience at Baidu influence your later views on China?

Amodei explained that he only worked there for one year, and that this experience at Baidu was not the main reason for the relevant stance he later formed. He is more concerned about the risks that increasingly powerful AI may bring, including mass surveillance and infringement of personal privacy.

Amodei's concern about AI risks has a long history. In 2021, he and his sister Daniela and others founded Anthropic precisely in the hope of building a company that places greater emphasis on AI safety.

In October 2024, Amodei published the famous long article "Machines of Loving Grace," depicting the enormous progress AI could bring in fields such as medicine, the economy, and scientific research, while also warning that advanced AI could be used for dangerous purposes such as surveillance and information manipulation.

By early 2025, with the sudden rise of DeepSeek, these discussions that had originally leaned toward long-term risks also gained a more realistic target.

In January, DeepSeek-R1 caused a sensation in Silicon Valley. Amodei immediately published the article "On DeepSeek and Export Controls," acknowledging that DeepSeek had made technological progress but arguing that its cost advantage had been exaggerated by the outside world, that advanced chips remained important, and further emphasizing the necessity of restricting the spread of technology.

In February 2025, when he was interviewed by ChinaTalk, the host asked a rather direct question: China has about one-sixth of the world's population. On the one hand, you hope AI will benefit all humanity; on the other hand, you support restricting China's access to advanced AI capabilities. Aren't these two things contradictory?

Amodei believes there is no contradiction. He gave the example that AI services could be provided to China through an API for beneficial work such as drug development and energy production, while restricting high-risk uses such as military applications.

However, when it comes to Claude's actual usage policy, this line is drawn quite strictly.

On September 4, 2025, Anthropic announced a further expansion of service restrictions on unsupported regions.

Previously, Claude had already restricted users in mainland China and other regions from directly using the service. This time, Anthropic extended the restrictions to the ownership structure behind enterprises. Even if a company is registered overseas, as long as more than 50% of its equity is directly or indirectly held by enterprises in unsupported regions such as China, it likewise cannot use Claude.

This means Anthropic is no longer satisfied with checking where users are using Claude; it also wants to further investigate who actually controls the enterprise.

Then on February 23, Anthropic directly named DeepSeek, Moonshot AI and MiniMax, accusing the three Chinese AI companies of conducting more than 16 million interactions with Claude through about 24,000 fraudulent accounts, extracting model capabilities without authorization.

It should be noted that the above attribution and data all come from Anthropic's own disclosures.

In Anthropic's view, large-scale distillation could allow competitors to replicate some of Claude's capabilities at lower cost. Preventing model capabilities from being extracted concerns both the AI safety it emphasizes and its own commercial interests.

From restricting service regions to reviewing corporate ownership to upgrading its anti-distillation system, the defensive line Anthropic has built around Claude is becoming ever tighter. But for ordinary users in mainland China, when this grand security philosophy lands on them, it means only one thing:

Even if they are willing to pay, they still cannot use Claude in mainland China through official channels.

03

Over the years, the scale of Anthropic's account bans has become increasingly astonishing.

In July this year, Anthropic's published transparency report showed that in the first half of 2026 alone, the company banned 11.4 million accounts, an average of about 63,000 per day, nearly eight times the figure for the second half of 2025.

During the same period, users submitted a total of 398,000 appeals, and in the end only 42,000 overturned the original penalty, a rate of about 10.6%.

However, no matter how strict Anthropic's restrictions are, they have not made Claude disappear from the Chinese-language internet. The stories about Claude not recognizing Chinese people have even become a meme.

For example, a rather absurd chat screenshot has recently been circulating online.

There are many similar meme images, and even the "compensatory leave" familiar to Chinese people can become a reason for Claude to ban an account.

Of course, these are all netizens' parodies and cannot be taken as real account-ban incidents.

But for domestic users who have studied network routes and computer time zones, and who even dare not chat with Claude in Chinese, this kind of joke is somewhat piercing.

Jokes aside, Claude has indeed not shut out all mainland Chinese users.

Although there are many account-ban experiences online, many domestic users have used Claude for several years and their accounts remain safe to this day.

In October this year, V2EX user vislearn shared a rather unconventional set of "account preservation tips": guard against nothing at all.

His computer time zone was Shanghai, his system language was Chinese, and Claude's memory even directly stated that "the user is from mainland China," yet his three accounts were still working fine.

He even once cursed Anthropic, and Claude merely deleted that conversation without banning the account.

Some users have also found that if Claude is used only as an ordinary chat assistant and does not involve complex technical tasks, it seems less likely to trigger an account ban.

But then again, if it is only for chatting, why insist on using Claude?

For many domestic developers, Claude's truly irreplaceable appeal is still its performance on complex tasks such as programming and agents. Claude Code in particular has become a tool that many developers cannot do without in their daily work.

Yet programming, agent and complex reasoning capabilities are precisely the core assets in which Anthropic has invested a large amount of R&D resources and strives to prevent competitors from replicating through large-scale distillation.

Although ordinary developers using Claude Code normally and large-scale extraction of model capabilities are two different things, this has also put Anthropic in a rather delicate dilemma: the harder it tries to protect Claude's technological advantages, the more these capabilities attract users who are willing to pay for them and even find ways around regional restrictions.

On October 8, Claude also quietly added simplified Chinese and traditional Chinese interfaces, and the payment page even briefly showed a UnionPay option.

Mainland China is still not on the official service list, but Claude has already begun to provide an increasingly complete Chinese-language experience.

No wonder some netizens could not help joking: Has Anthropic finally remembered to collect money from Chinese users, or is it preparing to lure people in and then ban them?

From Claude's first launch in 2023 to now, more than three years later, mainland Chinese users still have not seen an official opening.

The memes about its love of banning accounts, however, have been played with first.