Point system Claude uses to ban accounts

A social-media thread relays what its author describes as a Chinese community’s reverse engineering of Claude’s account-ban system. The reported model assigns accounts an initial score, then raises or lowers it according to registration, payment, usage, device, and network behavior; accounts that fall below an enforcement threshold are banned, while those nearing it may enter a review period marked by sign-in problems. The thread says scores recover slowly and older accounts have more tolerance. These are presented as community findings, not as a system confirmed by Anthropic.

The alleged risk factors center on account consistency and signals of legitimacy. A clean, stable IP and using one device are said to help, while frequent IP changes, multiple devices, rapid logins, and heavy concurrent use hurt. The thread also claims that mobile or desktop apps and enabled memory improve standing; immediate subscriptions, starting with the top-tier Max plan, virtual cards, and exhausting usage limits can count against an account. It recommends delaying paid plans, using a real foreign bank card or Google Play, and spreading usage across accounts. Linux and extensive use of Fable are also alleged to carry penalties, whereas location settings and telling Claude one’s nationality supposedly do not. The practical focus is on avoiding enforcement by shaping account behavior, though the scoring rules and claimed effects are not independently substantiated in the material.