Anthropic's October wave of Claude suspensions: what happened and how to protect your work

On October 1, a lot of Claude users started their morning with an email from Anthropic. The safety team said it had detected "suspicious signals" on the account that pointed to a violation of the usage policy, so access had been restricted. Two options were offered: file an appeal, or export the chat history and delete the account.

The emails went out in Hong Kong and in Russia on the same day, and they didn't discriminate. Free accounts, Pro subscribers and people who had paid $200 a month for Max for over a year all got the same message. Seniority didn't help, money didn't help, job titles didn't help. Department heads at large tech companies, for whom Claude was a daily work tool, were locked out too.

Here's what happened, why it isn't a random glitch but part of a course Anthropic has held since spring, and what to do if your work depends on Claude.

What exactly happened

The wave started on October 1 and spilled onto social media almost immediately. The story is the same for everyone: months or even years of using Claude without trouble, then one day an "Account suspended" screen instead of the chat.

In Hong Kong, most people had been connecting through a VPN exiting in Singapore, a country Anthropic officially supports, and paying with foreign cards. One user said he had worked that way through a popular commercial VPN for six months before the email arrived, and his first move was to rescue his chat history.

Elsewhere the setups were more elaborate. Some people used their own virtual server in Finland instead of a mass-market VPN. Some had stayed on the same foreign IP address for more than a year without ever hopping around. Neither helped.

Anthropic hasn't explained anything officially. No statement, no post about new rules, just emails with the same generic wording about "suspicious signals".

Why this isn't an accident

Anthropic publishes a list of countries where Claude is officially available. Hong Kong, mainland China, Macau and Russia aren't on it, while Singapore, Japan and Taiwan are. For a long time, using Claude from outside that list sat in a gray zone: it worked, it just wasn't supported. The company looked the other way for years, and in 2026 it started closing that zone methodically.

Look at the timeline.

In April, Anthropic introduced identity verification. In "some scenarios", Claude may ask for a passport or driver's license, a physical document in hand rather than a scan or photo, plus a live selfie. The check is run by an outside verification provider used by banks and fintech companies. The help article says plainly that accounts created from an unsupported region can be suspended, and that passing verification doesn't protect you from that.

In May, several hundred Russian accounts were shut off without warning, ordinary users and entrepreneurs whose workflows ran on Claude alike. No reason was given then either.

Hong Kong had been tightened even earlier. In the fall of 2025, Anthropic stopped serving companies that are more than 50% owned by Chinese capital. Later, major banks cut their Hong Kong staff off from Anthropic models on their own, without waiting for trouble.

And now October: a wave in two regions on the same day, following the same template. That looks more like a system than a one-off cleanup.

How Anthropic spots users it doesn't want

The company doesn't disclose its criteria, and that's understandable: the moment it did, people would start working around them. But from the stories of affected users and the analysis of people who study this, a few signals emerge that seem to work together.

Data center IP addresses. Nearly every VPN and VPS connects to the internet from a data center, not a home network. Those addresses have long been collected into databases and are easy to flag. That's an unpleasant conclusion for anyone who thought they were smarter than mass-market VPN users: your own server in Helsinki also sits in a data center. A fixed IP doesn't help if the address itself gives away the hosting provider.

Payment details. Virtual cards, a card from one country with an address in another, paying through intermediaries. For an anti-fraud system, these are classic red flags. They matter for $200 Max subscribers as much as anyone, arguably more: an expensive subscription with an odd payment trail draws extra attention.

Country hopping. Logging in from the Netherlands yesterday, Singapore today, Germany tomorrow. Real people don't travel like that, and the system notices.

What you write. Some experts are convinced this is checked too: if someone has written in one language for years and their browser lives in one time zone, an IP address from another continent convinces no one. How much weight this signal carries, nobody outside the company knows.

The takeaway is simple: there's no magic setting that guarantees an account stays safe. You can remove one signal, but the others stay. And if the system still has doubts, identity verification kicks in, which you can't pass from a region Anthropic doesn't serve.

If your account has already been suspended

Don't panic, but don't wait either. Here's what makes sense.

Export your data first. The suspension email offers a history export, and that's the first thing to do. Conversations, projects, uploaded documents, the long instructions you polished for months will be useful wherever you continue. Don't put it off: the terms for accessing a suspended account can change at any moment.

You can appeal, but don't count on it. Anthropic has an appeal form, and when a suspension really is a mistake, it's sometimes lifted. If the reason is your region, the chances are slim: on paper there is a violation, so there's nothing to argue about.

Don't create a new account with the same setup. Same card, same phone number, same VPN server, and the new account will most likely be linked to the old one quickly. Then you lose the money for the new subscription too.

Keep your working context on your side. System prompts, project descriptions and prompt templates are better kept in ordinary files, not only inside someone else's service. That lesson applies wherever you work next.

How to keep your work from depending on one company

The model hasn't gone anywhere. The real problem the October wave exposed is different: when everything you've built lives in one personal account with one vendor, a single email can take it all away.

ChatPlus is set up differently. You don't need an Anthropic account: you sign up with us using Google or email, and we connect to the Claude models ourselves, through the API. Your conversation history lives in your ChatPlus account, not in a service that might one day send you a note about "suspicious signals".

The full Claude lineup is here:

For most tasks, Claude Sonnet 5 is plenty: a good balance of quality and speed for both writing and code. For something big, like a project audit, a long document or serious analysis, go with Claude Opus 5.5. It's Anthropic's new flagship and currently tops the independent rankings.

Moving your work over is easy. You can upload the exported Claude archive to ChatPlus as is: chats, projects with their instructions, and files appear in your history with their original dates, and you can continue any conversation where you left off. We walk through it step by step in a separate guide. It's the same model, so the answers will feel familiar.

There's one more advantage that matters especially after news like this: you don't depend on one company. Claude, GPT, Gemini, DeepSeek and Grok live in one window, and you can switch models in the middle of a conversation. If something happens with one provider, your work doesn't stop.

An honest caveat: ChatPlus isn't a copy of claude.ai. The interface is our own, and Anthropic's own tools such as Claude Code for the terminal are separate products we don't replace. But the Claude models themselves work here in a regular chat.

The short version

On October 1, Anthropic launched a new wave of Claude suspensions in Hong Kong and Russia, regions it doesn't officially serve. Free users and $200-a-month Max subscribers alike lost access, and private servers or a fixed foreign IP didn't help. It isn't a one-off: in spring the company introduced ID checks, in May it already shut off accounts, and the direction is clear. If your account was suspended, export your history first and don't rush into an appeal or a new account. And whatever you use, keep your work somewhere that doesn't hinge on a single vendor: in ChatPlus, the whole Claude lineup from Haiku to Opus 5.5 sits next to GPT, Gemini and the rest, with your history in your own account.