A crazy blog.
Rohan Paul Twitter · Rohan Paul (@rohanpaul_ai) · 2026-06-28
Chinese developers are accessing Claude through gray-market proxy services that resell API tokens at 5-10% of official prices by aggregating farmed accounts, bypassing Anthropic's identity verification, and potentially monetizing private user data including code and reasoning traces.
Extraction
Topics: api-abusellm-securitygray-market-aichina-ai-accessdata-privacy
Claims
- Transfer station proxies in China resell Claude API access at 5-10% of official prices by aggregating farmed accounts and routing requests through overseas infrastructure.
- Transfer stations collect Claude accounts through free credit abuse, stolen payment cards, subscription splitting, and overseas verification workarounds.
- Proxies may silently substitute weaker models for Claude Opus, inflate token counts, and store or sell user prompts, code, and reasoning traces.
- This proxy architecture defeats Anthropic's KYC and account-banning systems because the company only sees the proxy, not the real end user.
- User private data including code, business tool calls, and reasoning traces may be stored by transfer station operators and monetized as training data.
Key quotes
The user thinks they are buying cheap inference, but the proxy may swap Opus for weaker models, inflate token use, or store private code, tool calls, reasoning traces, and business data.
This breaks a core assumption behind KYC, account bans, and abuse monitoring: the AI company sees the proxy, not the real person, so banning one account leaves the upstream supply chain alive.
The cheap price comes from account farming, free-credit abuse, resale of unused quota, subscription splitting, possible stolen cards, and a darker trade where user prompts and outputs become training data.