跪拜 Guibai
← All articles
Backend · Artificial Intelligence · Developers

Anthropic's Claude Code Ban Wave Hits Paying Users, Appeal Link Loops to Nowhere

By cxuanAI ·
Read original on juejin.cn ↗ Google Translate ↗ Alt translation

A paid SaaS tool that can terminate accounts without warning and offers no functioning recourse erodes the trust required to build production workflows on top of it. Developers betting their toolchains on Claude Code now face a single-point-of-failure risk that has nothing to do with model quality.

Summary

A wave of Claude Code account bans swept through the user base this week, catching App Store subscribers who had been using the service without issues for months. Users received termination emails with no prior risk warning, and the bans hit prominent Chinese developers including tech blogger Ruan Yifeng and entrepreneur Chi Jianqiang.

The appeal process is broken at a fundamental level. The termination email contains an appeal link that redirects to the Claude Code web interface rather than an actual appeals form. When users report this as a bug, support tells them to return to the same broken appeal page, creating a Kafkaesque loop: appeal entry to web redirect to support response back to appeal entry.

At least one affected user has already canceled their subscription in frustration, questioning whether Anthropic will even refund the remaining subscription period. The incident exposes a gap between Anthropic's infrastructure-grade product capabilities and its amateurish approach to user relations and trust management.

Takeaways
Anthropic sent account termination emails to Claude Code users with no prior risk warning, including subscribers who paid through the App Store and had clean usage histories.
Prominent Chinese developers Ruan Yifeng and Chi Jianqiang were among those banned, and both publicly criticized the move.
The appeal link embedded in the termination email redirects to the Claude Code web interface rather than an actual appeals page.
Reporting the broken appeal link as a bug resulted in support instructing users to return to the same non-functional appeal page.
At least one affected user immediately canceled their subscription and is waiting to see whether Anthropic refunds the remaining period.
Conclusions

Anthropic's ban mechanism appears to lack any graduated enforcement — no warning, no temporary restriction, just immediate account termination.

A broken appeal process that support cannot or will not bypass suggests the company has not invested in the operational infrastructure to match its AI product ambitions.

The gap between Anthropic's technical sophistication and its customer operations is wide enough that even high-profile, long-term subscribers get caught in it with no path to resolution.

Mass bans without explanation create a chilling effect: developers cannot assess whether their own usage patterns put them at risk, making the tool unreliable for production planning.

Subscription revenue from banned accounts creates a perverse incentive if refunds are not automatic, and the lack of clarity on refunds compounds the trust deficit.

Concepts & terms
Claude Code
Anthropic's AI-powered coding assistant, integrated into development environments to help write, debug, and refactor code using the Claude language model.
App Store subscription
A payment method where users subscribe to Claude Code through Apple's App Store billing system rather than directly through Anthropic, which adds Apple as an intermediary for refunds and account management.
From the discussion

The discussion splits between those who see the bans as a predictable consequence of violating terms and those who view them as a legal or political affront. One report claims the ban email arrived but access remains functional, undercutting the severity of the enforcement. A practical question about the necessity of the GUI over the CLI goes unanswered, while a final comment reframes the issue as a matter of national pride and circumvention.

Using a service after being explicitly denied access makes a ban self-inflicted.
The ban enforcement appears inconsistent, with access persisting despite a ban notification.
The bans are a straightforward matter of legal compliance in the provider's jurisdiction.
The bans are perceived as a targeted slight warranting a defiant, workaround-driven response.
A genuine use-case gap between CLI and GUI products is raised but not addressed.
Featured comments
时间掩盖

I received the ban email, but both the CLI and the desktop app are still working normally [polite smile]

一个大蜗牛

The CLI is enough for me. Under what circumstances do you guys absolutely need to use its GUI product?

兔子爱吃的萝卜

Why is no one talking about backbone at a time like this? When others find ways to ban us, we find ways to get around it.

See top comments, translated →
Source: juejin.cn ↗ Google Translate ↗ Backup ↗