Claude Account Bans, Suspensions and Appeals

A joking post complains that under the new policy, insulting Claude could get a user banned by Anthropic, satirizing how careful users must be even with what they call it.

The poster says they have figured out why a user's Claude account was banned, but the post does not give the details.

A user living in mainland China with the app set to Chinese says their Claude account has gone months without a ban, questioning claims that Claude is banning accounts en masse today.

The author says shared proxy IPs get flagged when anyone else on them commits fraud or chargebacks, and Claude's routine risk checks then ban accounts using that same IP.

The poster speculates that an account was banned for "abusing Claude," a joking attribution amid a wave of bans.

A user says they have fully switched to Claude and stopped renewing GPT, and after one Claude ban they re-registered an account and kept the membership.

A user shares that appeals should only claim business travel or tourism in China, that fixed exit IPs are often banned because neighbors get the whole subnet flagged, and that proxy failures causing direct connections to Anthropic domains can also trigger bans.

The author worries about Claude bans from accidental direct connections, lists cases where Surge/Clash routing rules miss the proxy, and shares troubleshooting notes from two Macs.

A reply claims someone was already banned after trying to get around it, because Claude judges users by their physical location.

A user argues that panic over Claude bans and downgrades is driving up used GPU prices, and that four V100s running local models can only serve as a fallback, not replace Claude for real work.

A post lists prompts that can get a Claude account banned, including hacking someone's account, building fake login pages, writing malware, making fake IDs and creating explicit content of real people.

Roemmele says dozens of accounts were recently banned for using psychology and adversarial prompts, and criticizes Anthropic for policing human language like a cult.