Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)
ReleaseAPI / Product · ActionableHabr AI→ original

Claude bans accounts right after payment: that happened in 22 of 101 cases the author analyzed

A Habr author who was himself banned on the Max plan collected 101 stories of Claude account bans from Habr, Hacker News and GitHub. The most common pattern, in 22 stories, was an account disappearing within minutes or a couple of days of a payment, renewal or plan upgrade, not because of a VPN. Among the 50 stories that named a plan, two-thirds involved Max. His advice: keep prompts and CLAUDE.md in git and run automation through an API key.

AI-processed from Habr AI; edited by Hamidun News
Claude bans accounts right after payment: that happened in 22 of 101 cases the author analyzed
Source: Habr AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceHabr AI ↗
Editor-in-Chief
Key metric #1User received a ban email and a receipt within one minute
Key metric #2Max 20x was named in 18 of the 50 stories that mentioned a plan
Key metric #3In 13 stories, people ran Claude through scripts and third-party tools

Imagine this: you click "Pay" for a Claude subscription, the money is charged, and a minute later your account is blocked and all your chats are gone. That is exactly what happened to the author of a Habr analysis: he was banned on the Max plan, Claude Code suddenly stopped working one day, and a blocking email was waiting in his inbox. The money was not refunded. He then collected 101 stories of such bans and found that the most common case, in 22 of 101, was a ban arriving at the moment of payment. Not a VPN, as he had expected.

He pulled the stories from three places: comments on Habr (mostly from Russia and the CIS), Hacker News threads, and issues in the Claude Code repository on GitHub. He counted only first-person accounts, and if someone complained in several places, he counted them once. This is not Anthropic's statistics: each person names the cause they believe in. But when the same storyline repeats across dozens of accounts, it is hard to dismiss.

The most dangerous moment is payment

You buy a subscription, upgrade to a more expensive plan or renew an old one, and within a minute (two days at most) the account is gone. There are 22 such stories out of 101, more than any other identifiable cause. A quarter of people have no idea what they were banned for.

A Hacker News case is telling: a user signed up, purchased Pro and clicked "Pay". He was immediately logged out. Two emails arrived within a minute: one about a policy violation and one receipt, with the receipt arriving later. On Habr, one commenter paid for a subscription with a virtual card from a crypto exchange and promptly got both a ban and a refund. His friend tried an Armenian card belonging to someone else and got the same result.

Anthropic does not explain the reasons. The author's conclusion: during payment, the service simultaneously learns the card's country, the bank, the payer's address and the address you are paying from, and checks this against what it already knows about the account. This is his hypothesis, not a confirmed fact.

The more expensive the plan, the higher the chance of being removed

A plan is named in 50 stories. In 18 of them it is Max 20x at $200 per month, and together with the other Max versions that makes up two-thirds. Pro came up eight times, the free plan four times.

The author links this to the hunt for account "farms." In February, Anthropic reported that the Chinese companies DeepSeek, Moonshot and MiniMax set up about 24 thousand fake accounts, ran more than 16 million requests through Claude and used the answers to train their own models. Such accounts are now caught by their behavior. An active Max 20x user looks similar: they stay at their limits around the clock, run Claude Code in several windows and leave agents running overnight.

Another 13 stories involve scripts and third-party programs. One person ran Claude Code from their own script and was removed almost immediately. Another entered subscription details into the JetBrains assistant, saw that an API key was required, closed the tab, but 20 minutes later received a ban email. Anthropic's terms explicitly prohibit access by "automated or non-human means" that does not go through an API key, and since April 4 the subscription no longer covers third-party tools such as OpenClaw.

What about VPNs? Nearly every third story on Habr complains about the network, and Anthropic's help page names creating an account from an unsupported region. But the picture is contradictory: one person was banned while using a residential proxy, while another switched five countries in a day through free VPNs with no consequences. It seems the address itself is not what matters, but a sharp change in the account's "identity": for example, logging in from a different browser through a proxy or reconnecting the VPN. Half of the Habr stories happened after September 20, so the checks appear to have been tightened.

There are also unexpected causes. A 15-year-old developer who paid for Max himself had his account disabled with the wording "signs that the account is being used by a child." It can be restored through age verification with the service Yoti. And an adult who asked for a fact-check on the USSR between the world wars was asked to confirm their age the next day. Creating a new account after a ban is risky: one person who paid for a new account with the same card got the same ban, although a couple of users are working calmly on new accounts.

What to do in practice

There is no magic setting that protects against a ban. But some steps are possible. First, secure your data immediately: according to Anthropic's help page, even a banned user can log in to claude.ai and export their chats. However, exports can be truncated: one Hacker News user, after more than two years on higher-tier plans, received a 235-byte JSON containing only a name, phone number and time zone. The author's chats also disappeared, and only what the agents did through the API survived.

Therefore, developers should keep prompts, CLAUDE.md and their work in git rather than only in chats, export conversations in advance and run autonomous workflows through an API key. It is equally sensible to keep a backup model to switch to within a day if the account disappears at the worst possible moment.

Story

Key Questions Answered

What are the key details on The most dangerous moment is payment?

You buy a subscription, upgrade to a more expensive plan or renew an old one, and within a minute (two days at most) the account is gone. There are 22 such stories out of 101, more than any other identifiable cause. A quarter of people have no idea what they were banned for. A Hacker News case is telling: a user signed up, purchased Pro and clicked "Pay". He was immediately logged out. Two emails arrived within a minute: one about a policy violation and one receipt, with the receipt arriving later.

What are the key details on The more expensive the plan, the higher the chance of being removed?

A plan is named in 50 stories. In 18 of them it is Max 20x at $200 per month, and together with the other Max versions that makes up two-thirds. Pro came up eight times, the free plan four times. The author links this to the hunt for account "farms." In February, Anthropic reported that the Chinese companies DeepSeek, Moonshot and MiniMax set up about 24 thousand fake accounts, ran more than 16 million requests through Claude and used the answers to train their own models.

What to do in practice?

There is no magic setting that protects against a ban. But some steps are possible. First, secure your data immediately: according to Anthropic's help page, even a banned user can log in to claude.ai and export their chats. However, exports can be truncated: one Hacker News user, after more than two years on higher-tier plans, received a 235-byte JSON containing only a name, phone number and time zone. The author's chats also disappeared, and only what the agents did through the API survived.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace · Nomad

Test Frontier LLMs & Autonomous Agents in One Workspace

Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.

Try Nomad AI →Launch Nomad Agent
What do you think?
Loading comments…