Claude Account Bans, Suspensions and Appeals

A user considering switching from CC Switch to Magpie asks whether connecting official Claude and Codex subscription accounts risks rate limits, verification prompts, or account bans, and wants long-term user reports.

The author's two-year-old Claude account was banned, and new accounts with fresh IPs, Gmail, US App Store cards or purchased accounts were banned instantly, concluding any setup triggers bans.

Reminds Claude users on proxies that refunds don't restore chat history, so export data periodically via Settings → Privacy → Export data.

A user says many people worry that paying for Claude Max will get their account banned, so they register several accounts and switch between them when tokens run out.

An image post asks why so many people are having their Claude accounts banned, reflecting how common the issue has become among users.

Post shares an image of unconventional, questionable tricks users try to keep their Claude accounts from being banned.

The author subscribed to Claude Pro with a new Google alt due to phone verification; in the July ban wave both the alt and the never-used main account were banned, and the main account was banned again after a week of free use in October.

A seller is offering pre-made Claude accounts with a ban warranty after two months of testing, priced high for stability, with Taobao and WeChat ordering plus a support group.

A user asks whether it's true that Anthropic recently loosened its ban rules and that many users' Claude accounts have been reinstated.

A user says a modified sub2api deployment on an overseas server, with CLI-aligned fingerprints, has kept their Claude Pro account alive for three weeks.

Elvis rights holders claim Anthropic trained on 274 songs and are seeking about $41 million; the poster jokes that part of the payout comes from their own banned account.

A replier who also got banned right after paying for membership suggests splitting spec-writing and rendering across models, using Cursor with GPT/Codex and Gemini for long prompts so one failure doesn't stop the pipeline.