Do not set up "AI relay proxy stations" to resell overseas large language model API access at discounted rates: operators have been criminally detained

"AI relay stations" are unauthorized proxy websites that batch-purchase overseas LLM API subscriptions and resell access tokens cheaply to domestic users. In Shanghai, an unauthorized relay operator was criminally detained for 37 days, subsequently released on bail while awaiting formal prosecution and confiscation of unlawful proceeds. National security agencies have issued warnings that many illicit relay proxies lack telecom licenses and unlawfully route domestic user prompts and sensitive data to overseas servers.

Zero financial outlay. The opportunity cost is giving up an easy, high-cash-flow side hustle. Legitimate enterprise model aggregation requires telecommunications operating permits, authorized corporate API contracts, and substantial compliance capital.

In June 2026, the Ministry of State Security (MSS) issued a formal security advisory regarding "AI Relay Stations." The advisory outlined that unauthorized relay stations act as unregulated intermediary proxy layers between end-users and official model providers. Many such proxy operations operate without basic telecommunications and data processing qualifications. They systematically bypass cross-border data transfer security assessments and statutory approval processes, transmitting user inputs and corporate proprietary data directly to foreign servers, while facilitating bypass of network access controls and licensing restrictions. In an investigative report by China Newsweek (June 2026), an operator of an unauthorized AI proxy portal published a public shutdown announcement in a technical community after being held in criminal detention by Shanghai police for 37 days for illicit API acquisition, followed by release on bail pending trial. The operator confirmed facing imminent criminal sentencing alongside restitution, disgorgement of illicit gains, and criminal fines. Unregulated proxies typically bulk-register overseas enterprise accounts or exploit leaked API tokens, routing traffic through foreign jump servers to undercut official pricing.

Grade C is assigned because this enforcement pattern is documented through investigative media reports, public regulatory advisories, and operator disclosures, without publicly released final court verdicts detailing formal statutory criminal charges. While legal literature suggests charges ranging from illegal business operations to computer information system offenses, operators face verified risks of felony detention. Lawful model aggregation requires commercial telecommunications value-added licenses (see Chapter 26, Rule 2), statutory generative AI security filings (see Chapter 11, Rule 17), and strict compliance with cross-border data transfer regulations (see Chapter 26, Rule 10). Risks to end-users utilizing unauthorized AI relays are covered in Chapter 14, Rule 10. Benefit magnitude is categorized as High based on avoiding criminal detention and asset forfeiture. Primary beneficiary is yourself.

国家安全部 (2026). 「AI中转站」，风险要防范. https://www.szzg.gov.cn/2026/xwzx/szkx/202606/t20260608_5331487.htm（数字中国建设峰会官网转载国家安全部微信公众号）；中国新闻周刊 (2026-06-22). AI中转站，面临整治「风暴」（记者孟倩）. https://www.chinanews.com/cj/2026/06-22/10644787.shtml（网页存档 https://web.archive.org/web/20261005042903/https://www.chinanews.com/cj/2026/06-22/10644787.shtml）