Claude Account Bans, Suspensions and Appeals

A blogger reads Anthropic's anti-distillation post as a guide to avoiding bans: don't elicit or tamper with chain-of-thought, keep registration and payment channels clean, and don't mix domestic models with Claude in one agent environment.

Anthropic says distillation attacks on Claude now yield 90-99% less useful data, citing limits on passing reasoning across contexts, a two-stage classifier, and bans plus charges for repeat extractors, with tens of thousands of fake accounts banned.

A user who has used Claude since early this year without interruption lists anti-ban practices: a long-term personal Gmail, a real US credit card and address, a US static IP, and a fingerprint browser.

Anthropic says companies secretly used Claude to distill their own models and has banned tens of thousands of related accounts, often within two hours, with some legitimate developer connections also affected.

The poster says that in China even a hard-won subscription does not guarantee the Claude account will not be banned tomorrow, and that stable use is all down to luck.

A post says that using rea with Codex and Claude may lead to official account bans, warning users to be careful.

Anthropic says its anti-distillation measures cut the reasoning tokens distillers can obtain by 10–100x, banned tens of thousands of accounts (most within two hours of starting), and named Zhipu's GLM-5.3 and Moonshot's Kimi; the poster notes these are one-sided claims.

Anthropic says distillers can now scrape 90%–99% less reasoning data, has banned tens of thousands of accounts (many within two hours of starting), and named Zhipu and Kimi; blocked requests are still billed and legitimate developers may need to change integrations.

Anthropic published a post accusing Zhipu and Moonshot of distilling Claude's capabilities, saying it uses classifiers to block extraction, has banned tens of thousands of accounts, and will charge for blocked requests from September 24.

Anthropic says its new anti-distillation mechanism cut the data scraped from Claude by 90%–99% and banned tens of thousands of distillation accounts, directly naming Zhipu's GLM-5.3 as having distilled Claude.

A long read breaks down Anthropic's anti-distillation blog: reasoning summaries, classifiers, tens of thousands of fake accounts banned, and the cost of false positives on legitimate requests that are still charged.

A post walks through avoiding a Claude ban end to end, from IP and an aged Google account to clearing Skill residue and a Singapore payment card.