A Reddit user was working in Claude Code on personal projects. Then the service asked him to verify his identity. He submitted his Russian international passport — and immediately saw that his account had been blocked. According to him, he lives in New Zealand on a valid four-year visa, and the money for the next month of his subscription was charged two days before the ban.
Anthropic conducts verification through Persona: a photo ID plus a selfie. Russia is absent from the list of supported countries. The company itself lists creating an account from an unsupported region among the grounds for restricting access. Chat history doesn't necessarily disappear upon blocking: Anthropic allows data export, though the amount depends on the reason. This archive cannot be automatically transferred to another personal Claude account. The case is analyzed by SpeShu.AI on Habr.
The source doesn't explain why the check was requested from this particular person — and that's the main hole in the story. The sequence "submitted a passport → got banned" doesn't become a rule on its own. Our assessment: the combination of "identity check plus geography" has long loomed over Russian-speaking Claude users; it just hadn't triggered so visibly until now.
The real news here isn't just one ban. A chat with a model has become a working archive: rejected libraries, client feedback, and already-debugged errors all settle into the conversation. We believe that's exactly why a block hurts more than it seems from the outside. A blank new chat knows none of this, and no one moved the deadline.
Intermediaries win. The Habr article was published in the SpeShu.AI blog — a company that sells precisely a workaround for this block: access to Anthropic models through its own interface, payment in rubles, a Russian IP is no grounds for a ban. The promo here is honest, but it's still promo. Users lose: money has been charged, the tool is inaccessible, and the work context is locked inside the account.
There is no official confirmation: the reason for this particular ban hasn't been publicly established, and so far it's a single voice from Reddit, secondhand. The retelling itself lives in the blog of a company that sells workaround access to Claude — it benefits from the case looking like a rule. The text also carries the date "October 29, 2026"; it looks like a typo, but there's no way to verify. Worth checking yourself: whether Anthropic has written verification terms for people who live in a supported country but hold an unsupported passport.
Forecast: over the coming quarter I expect a few more public cases of "submitted a passport — lost the account": the verification scheme is the same for everyone. If they don't appear, it means the checks trigger selectively, not on the mere fact of a Russian document.
Disclaimer: not officially confirmed, single source.
Source: SpeShu.AI on Habr
Source: habr.com · post in Telegram