China’s Ministry of Public Security has launched a new anti-scam service built around large language models, multimodal AI, and agent technologies, expanding the country’s use of artificial intelligence in public-facing fraud prevention. The National Anti-Fraud AI application went live on September 6 and is available through major Chinese app stores as well as WeChat and Alipay mini-programs.
Official reporting identifies the Shanghai Public Security Bureau as the developer and says the Ministry of Public Security’s Criminal Investigation Bureau directed the project. Its launch arrives after years of Chinese efforts to combine public warnings, platform enforcement, and financial controls against telecom and online fraud. The new element is an effort to put generative and agent-style interfaces directly into a public anti-fraud tool.
That does not mean the app has already proved it can reduce fraud losses. The government has described it as a prevention and education service, not as a validated detection system with published outcome data. Its importance lies in the design choice: make AI a conversational front end for people deciding whether a message, payment request, or online offer might be dangerous.
A public interface for AI-assisted risk checks
According to the official description, the app provides an AI question-and-answer function intended to help users assess scam-related risks. It also offers an anti-fraud media section with real-time warnings and a digital anti-fraud dictionary that can read explanations aloud. The technical stack reportedly combines large language models, multimodal models, and AI-agent capabilities.
In everyday use, that could mean a person describes an unfamiliar investment pitch, employment offer, phone call, or online transaction and receives guidance about common warning signs. A multimodal system could potentially process more than text, while an agent-style system may help organize a response or surface relevant educational material. The authorities have not published a full technical design, so the app’s detailed capabilities and limits remain unclear.
The government’s approach is notable because it treats anti-fraud assistance as a continuous service rather than a static awareness campaign. Users can access it where they already communicate and pay, through WeChat and Alipay mini-programs. That distribution could improve convenience, but it also means the service will need to provide reliable, understandable advice in situations where panic, urgency, or social pressure are central to the scam.
China has pursued AI governance alongside AI deployment. EastFrontier covered a months-long campaign against AI misuse, deepfakes, fraud, and data poisoning earlier this year. The new app represents a different layer of policy: using AI not only as something to regulate, but also as a tool for public protection.
The app is separate from China’s older fraud tools
The National Anti-Fraud AI app should not be confused with China’s older National Anti-Fraud Center app, which is widely associated with public reporting, warnings, and prevention features. The September 6 launch is described as an AI-specific service developed by the Shanghai police under national guidance.
That distinction matters because the public conversation around digital fraud tools often mixes separate applications, databases, and enforcement systems. The older app’s experience may provide a distribution and awareness foundation, but it does not establish what the new AI service can do or how it handles user information.
The official release does not set out detailed privacy rules, retention periods, data-sharing arrangements, or the conditions under which a user’s query might be reviewed by a person. Those omissions should not be filled with assumptions. An anti-fraud tool can be useful while still raising questions about transparency, false positives, and how sensitive financial or personal information is handled.
The risk of incorrect advice is also real. A conversational model may help a user identify familiar scam patterns, but it cannot guarantee that every suspicious message is fraudulent or that every apparently normal transaction is safe. The practical standard will be whether the app makes people more likely to pause, verify, and use established reporting channels when a situation looks risky.
China is making AI part of public-service delivery
The launch fits a wider pattern in China’s public-sector AI strategy. Authorities are testing AI in customer service, health, education, transport, and local administration, often through applications that combine official information with a conversational interface. In September, China’s new human-AI customer-service standard took effect, emphasizing clear handoffs and safeguards when automated systems handle consequential tasks.
Anti-fraud work creates an even higher bar. A poor answer about a purchase may be inconvenient. A poor answer about a scam can have financial consequences. For that reason, the most valuable function may not be fully automated judgment, but structured guidance that helps a user find the right warning, official contact point, or reporting route quickly.
Authorities have framed the service as part of a broader modernization of fraud prevention. That objective is understandable in a market where scams change quickly and can use manipulated images, voice cloning, fake investment dashboards, and social-engineering tactics. AI systems may help classify those patterns at greater scale, but criminals can also use AI to make their operations more persuasive.
The central test will be whether the app can keep pace with that contest. Its first contribution may be modest: give citizens an easier way to ask questions before they transfer money or share information. Over time, officials may add functions and publish more technical detail. Until then, the September 6 launch is best understood as a new public-service experiment, not as proof that AI has solved online fraud.