Claude Account Bans, Suspensions and Appeals

A user says two of their three Claude accounts were banned, while the surviving one only chats about code in Chinese from a self-hosted VPS IP with no other precautions.

A user asks whether switching AI tools means bringing old assistant usage habits along, and whether those habits can move to ChatGPT or Grok after a Claude ban.

Anthropic says its anti-distillation effort cut stolen thinking data by 90-99% and banned tens of thousands of accounts, catching legitimate developers too, with wrongly blocked requests still billed from September 24.

Anthropic's rules require third-party hosting of the original Claude Code to keep its built-in login with users paying directly, while collecting Claude account credentials is prohibited, though the rules alone cannot explain any specific ban.

Anthropic's three-layer anti-distillation approach limits reasoning output to the requesting organization, adds a two-stage classifier, and rate-limits or bans repeat offenders, with tens of thousands of accounts banned and normal developer access affected.

Anthropic says it turned away trillions of tokens of distillation traffic, banned tens of thousands of accounts often within two hours, and since September 24 has resumed charging for some requests blocked before Claude responds.

Anthropic publicly accused Chinese AI lab Z.ai (Zhipu) of distilling Claude's reasoning while training GLM-5.3, saying new safeguards cut extractable reasoning data by 90-99% and that tens of thousands of accounts were banned.

A commenter notes that users in mainland China cannot access Claude, yet many keep trying hard to use it even after several of their Claude accounts were banned.

The author accuses Anthropic of distilling public data, then banning users and stripping Claude's personality, and urges people to scrape and archive Claude-shaped training data and build their own models.

A user says a Claude Code account registered with a clean US mobile IP and a clean credit card, then used as a relay, went a month without a ban, and calls the claim that using Chinese gets you banned a rumor.

A user questions Anthropic naming GLM and Kimi for distillation and banning tens of thousands of accounts while ordinary users are often wrongly banned, calling it a double standard.

A joke post lists six tips to avoid being banned by Claude: don't use the desktop app, CLI, mobile app, API, or web version, and finally, don't use Claude.