--- title: "Fake Records, Real Sentences: China's Courts Meet AI-Forged Medical Extortion, The Take Inside China" date: 2026-09-05 originalDate: 2026-09-05 originalTitle: "AI伪造病历索赔？已有人获刑" issue: "Issue 7" description: "China is sentencing AI-forged medical extortion: four months' detention for fake gastroenteritis records; a ¥1,709 mapo-tofu claim now faces charges." tags: - "AI misuse" - "deepfakes" - "extortion" - "courts" - "food delivery" - "insurance fraud" sources: - label: "Guangming Daily via People's Daily Health: Yang case, extortion conviction with AI-forged records (2026-07-11)" url: "https://health.people.com.cn/n1/2026/0711/c14739-40758469.html" - label: "KNEWS via Sina Finance: Xuhui case, ¥1,709 mapo-tofu claim, criminal compulsory measures (2026-07-13)" url: "https://finance.sina.com.cn/jjxw/2026-07-13/doc-inihsckv7471711.shtml" - label: "Beijing Daily via Tencent: MPS names 'injury broker' insurance-fraud ring (2026-05-22)" url: "https://news.qq.com/rain/a/20260522A007Q200" faq: - q: "What did Yang do, and what was the sentence?" a: "Between November 2025 and January 2026, Yang dined at or ordered from multiple restaurants, then used AI tools to fabricate gastroenteritis medical records, diagnosis certificates, e-invoices and even photos of vomit for himself and his girlfriend, threatening complaints to regulators, the 12315 hotline or bad reviews unless restaurants paid 'medical fees.' He obtained ¥2,500 with another ¥7,000+ attempted, was arrested in January 2026, and was convicted of extortion (敲诈勒索罪): four months' criminal detention suspended for four months, plus a ¥2,000 fine." - q: "What is the Shanghai Xuhui case?" a: "Duan, physically located in another province about a thousand kilometers away, remotely ordered mapo tofu from a Xuhui district restaurant, claimed a wire in the food cut his tongue, and used AI to swap his name onto an injury photo scraped from the web. He collected ¥670, then returned the next day claiming the injury worsened and extracted another ¥1,039 — ¥1,709 in total, ¥3,000+ cumulatively. Xuhui police placed him under criminal compulsory measures on extortion suspicion; the case remains under investigation, so unlike Yang he has not been convicted." - q: "Why do these scams work?" a: "The merchant's weakness is structural: fear of complaints, platform intervention and rating damage makes a quick payout cheaper than a fight, and AI-generated invoices and diagnoses with hospital stamps are now nearly indistinguishable to the naked eye. What once required connections and forgery skills now takes a prompt." - q: "Is insurance the next front?" a: "The column's argument, and it has official backing: on May 20, 2026 the Ministry of Public Security publicly named an 'injury broker' ring — the Li case in Henan — that defrauded insurers of over ¥100 million in claims, with 11 arrested and insurance-fraud convictions in December 2025. That ring worked by altering real medical records; AI collapses the cost of the alteration itself, which is why the pressure is shifting to the verification end." --- Order a mapo tofu delivery, claim it made you sick, produce a hospital receipt, and collect ¥1,709 from the restaurant. The money arrived; the receipt was AI-generated in seconds; the "patient" is now in the criminal-justice system. Our WeChat column's take ran under the headline "AI-forged medical records for compensation? Someone has already been sentenced" (2026-09-05). This entry translates the take with the court record pinned underneath — including where "already sentenced" outruns the record for one of the two defendants. ## The numbers - Yang case, Nov 2025 – Jan 2026: AI-forged gastroenteritis records, diagnosis certificates, e-invoices and vomit photos for himself and his girlfriend; threats of regulator complaints, 12315 hotline reports or bad reviews; ¥2,500 extorted, ¥7,000+ attempted; arrested January 2026 (Guangming Daily via People's Daily Health, 2026-07-11) - Sentence: extortion (敲诈勒索罪) — four months' criminal detention, suspended for four months, ¥2,000 fine; the court found "no intent to defend rights, only illegal profit," leveraging merchants' fear of reputation damage and business-suspension checks - Duan case (Shanghai Xuhui): located ~1,000 km away in another province; remote mapo-tofu order; AI swapped his name onto a scraped "tongue cut by wire" photo; ¥670 first demand + ¥1,039 next-day escalation = ¥1,709; ¥3,000+ cumulative; criminal compulsory measures on extortion suspicion, case still under investigation (KNEWS via Sina Finance, 2026-07-13) - Backdrop: on 2026-05-20 the Ministry of Public Security named an "injury broker" insurance-fraud ring — the Li case, Henan, ¥100 million+ in fraudulent claims, 11 arrested, insurance-fraud convictions handed down December 2025 (Beijing Daily via Tencent, 2026-05-22) - **Figure arbitration:** the take says Yang's spree spanned "two months"; the official record says November 2025 through January 2026. And "someone has already been sentenced" fits Yang exactly — but Duan is at the compulsory-measures stage, pre-indictment. Yang's sentence is four months' criminal detention (拘役), not imprisonment (有期徒刑), and it is suspended: the durable price is the criminal record, not time served. ## The take inside China **Three fake receipts, three outcomes.** The take's framing, and the record supports it: Yang — convicted, sentenced. Duan — criminal compulsory measures, awaiting prosecution. A grill-restaurant customer who came in with AI-generated records — detained. Same method, escalating consequences, all within a single summer of enforcement. **Why it works on merchants.** The scam's engine is not the AI, the take argues, it is the merchant's calculus: fear of bad reviews, fear of platform intervention, fear of rating scores. A stamped-looking receipt pushed across the counter is cheaper to pay than to contest — minutes versus days. AI didn't invent that leverage; it industrialized the evidence. **The next front is insurance.** Restaurants are small money. The take points at the real pool: injury-claim insurance fraud, where "injury brokers" altering real medical records were already named by the Ministry of Public Security in May. Insurers' warnings run the same way — forged records mean denial, contract termination, and criminal liability. The take's closing line for readers tempted by the "business": the endgame is already written into case law. Don't touch it. ## What the Chinese take left out The actual sentences. "Already sentenced" carries a weight the record partly withholds: Yang got four months' criminal detention, suspended — a real conviction, but no prison time, and the total haul was ¥2,500. The deterrence story rests more on the criminal record and the pace of prosecutions than on the punishment's severity, a nuance the take skips. The take also does not name the statutes — the two cases are charged as extortion, while the insurance-fraud backdrop is a different crime family — and it leaves platform-side verification duties (who should be checking stamps and invoices at scale?) entirely unexamined. English-language coverage of both cases was essentially absent at press time; the conversation in English still runs on generic deepfake-fraud statistics rather than named, sentenced Chinese defendants. ## Why it matters outside Generative AI has collapsed the cost of forged evidence, and the first wave of criminal law's answer is visible here: not new statutes, but old extortion law applied to AI-forged claims, with sentences attached. Every review-driven, chargeback-prone marketplace — delivery, hospitality, e-commerce — faces the same asymmetry the take describes: verification costs more than payout, until a court changes the price. China's verdicts are among the first to pin AI-fabricated victimhood to criminal extortion, and the Ministry of Public Security's May naming of "injury brokers" signals where enforcement goes when the stakes move from ¥1,709 dinner claims to ¥100 million insurance pools. This is the second entry in our file on fabricated-evidence economics — it follows the AI-fabricated chat records that hijacked Weibo ([Issue 3](/watch/2026-08-30-ai-fabricated-chat-weibo-ban-inside-china/)) — and the pattern both share is the one to watch: the forgery is cheap, the verification is not, and the gap is being closed in court. ## Sources - [Guangming Daily via People's Daily Health: Yang case, extortion conviction with AI-forged records (2026-07-11)](https://health.people.com.cn/n1/2026/0711/c14739-40758469.html) - [KNEWS via Sina Finance: Xuhui case, ¥1,709 mapo-tofu claim, criminal compulsory measures (2026-07-13)](https://finance.sina.com.cn/jjxw/2026-07-13/doc-inihsckv7471711.shtml) - [Beijing Daily via Tencent: MPS names 'injury broker' insurance-fraud ring (2026-05-22)](https://news.qq.com/rain/a/20260522A007Q200) > **Provenance & disclosure.** Originally published in Chinese on our WeChat channel on 2026-09-05 ("AI伪造病历索赔？已有人获刑"); drafted with AI assistance under human editorial direction. Translated to English on 2026-09-05 (AI-assisted, human-reviewed). This entry goes beyond translation: Yang's sentence (four months' criminal detention, suspended, ¥2,000 fine), the ¥670 + ¥1,039 = ¥1,709 arithmetic in the Xuhui case, and the MPS naming of the injury-broker ring were each verified against the Guangming Daily, KNEWS and Beijing Daily source pages. Where the take's framing ("two months," "already sentenced" applied broadly) diverges from the record, the Figure arbitration line above flags it. This is translated commentary — not a SigPulse measurement. Our first-party measurements live in the [dispatches](/posts/) and the [/data/ ledger](/data/).