ChatGPT captures 18 corruption cases in New Taipei City! Civil servants from the Audit Department use an app to track 2 million judgment documents.

AI can not only choose potatoes, but also catch cheating cases and let civil servants upgrade independently, but the entire government needs to face up to the speed of AI and legislation. (Synopsis: Young people use ChatGPT as a life guide!) OpenAI CEO Sam Altman: Over 35 years old can't keep up at all) (Background added: Bloomberg: DeepSeek's strong rise, China is posing a "huge threat" to US AI supremacy) Who would have thought that such a serious-sounding event as tracking down government procurement fraud cases would be related to daily mobile phone photos and even free apps that clear mobile phone capacity? This is not a suspense novel plot, but a real story that takes place in Taiwan's audit department. According to an exclusive report by Tianxia Magazine, "Young civil servants rely on ChatGPT, free photo app. A young audit civil servant, with the help of AI tools, successfully caught a labor procurement fraud case of NT$150 million from the New Taipei City Government. Wu Xinmei, a young civil servant in the audit department. Like many people around us, her phone is crammed with photos of her family, especially the growth of her baby children. In order to solve the problem of insufficient capacity, she downloaded a free app called "Photos Clean", originally just to clean up duplicate photos, but never expected that this seemingly ordinary tool would later become her artifact to expose procurement fraud. Wu Xinmei, who joined the audit department four years ago, is one of more than 700 "government wallet guardians" in Taiwan, shouldering the heavy responsibility of supervising government procurement cases. Last year, she teamed up with Xu Zhongdao's senior auditor to conduct an in-depth investigation of New Taipei City's labor procurement cases such as maintenance, inspection and horticultural maintenance of public facilities in the past three years. The amount of such cases is not high, the number is huge, and the traditional manual inspection can be said to be a needle in a haystack. Facing 2 million verdicts with ChatGPT First, they are facing a huge number of judicial verdicts. In order to identify "high-risk" manufacturers with bad records but still be able to win bids, nearly 2 million verdicts must be screened. In the past, manual interpretation alone was almost an impossible task, Wu Xinmei told Tianxia Magazine, "I first asked ChatGPT to help me write a Python code." Through the crawler generated by AI, they quickly screened out the manufacturers who had been convicted by the court in the past ten years from 2 million verdicts, and the results showed that there were 6 government agencies and schools in New Taipei City alone, and a total of 18 procurement cases were among them. Xu Zhongdao explained that most of these problematic manufacturers were suspended by the court, so the administrative authorities did not receive formal notice and failed to be blacklisted in time, allowing them to continue to bid viciously for government oil. After the Audit Department has identified these 18 high-risk procurement cases, the next performance document review is the real big challenge. The documents of these 18 cases alone are piled up, and it is impossible to digest them in a limited time with only two manpower. That's when Wu Xinmei's day-to-day experience came in handy, and she thought of the app that helps her clean up her phone's photos, "Photos Clean," which detects and flags duplicate or similar images. This inspired her: manufacturers often attach photos before, during, and after construction as proof in their performance reports, and if there is a fake, it is likely that the same photo will be used to impersonate records from different locations and at different times. Therefore, Wu Xinmei took the performance photos of these procurement cases with her mobile phone for a whole working day, and imported them into "Photos Clean" for comparison. This test really made many "evil ghosts" hiding in the pile of documents appear! Through duplicate or highly similar photos flagged on the app, they quickly identified suspicious documents, and further investigation found that as many as 16 of the 18 procurement cases were suspected of forging documents, involving a total contract amount of more than $150 million. "Our purpose is not to find ballast," Li Shunbao, deputy auditor general of the audit department, said in an interview with Tianxia magazine, "but to tell the administrative authorities that by making good use of these new methods, they can actually find hidden problems more efficiently." After the audit report submitted by the audit department, the New Taipei municipal government not only transferred the manufacturer involved in the case to justice and actively recovered part of the money, but also showed a high willingness to learn, the New Taipei municipal government took the initiative to ask the audit department for help, and decided to incorporate automatic detection technologies such as images, faces, and locations into the future manufacturer's performance and acceptance process. This represents that this time the "AI fraud detection" model is expected to change from a case-by-case to a systematic fraud prevention mechanism. Comment: Where is the threshold for government AI? This fraud investigation initiated by the audit civil servants of the audit department can definitely brighten the eyes of the public, and the public service system can also use AI to break through the dilemma of insufficient resources, and everyone who has been in the public knows that there are always more things to solve than resources, and the resources have not yet arrived, and the KPIs at each point in time are pouring in again. Civil servants can only find their own resources, either to find contacts or to specialize in additional tools, and there are few open resources like AI models that "one person can use, everyone can use". The AI tools used by civil servants in the above story are not unattainable expensive programs, but free apps and common ChatGPT, which also greatly lowers the threshold for public AI adoption. But the key to the public is the examination of confidentiality, the trust of manufacturers and practical results. Who guarantees that the data of the passing AI will not be leaked? Can the Ministry guarantee that these AI tools are all in compliance with the Government Procurement Law? Secondly, the above story describes that the current government procurement law has loopholes in the management of "slow prosecution" manufacturers, which are ubiquitous in the public system, and if AI is added to cross-departmental systems to make it more efficient, then the end point of AI is, can civil servants be replaced by AI agents in large numbers? The current situation is that there is no formal AI legal regulation even for the private sector, so how long will it take for the government to become AI? Related reports The US Department of Commerce warns the world that "Huawei AI chips are not allowed": theft of US technology and violation of export controls will be punished Podcast essence: AI and robots cover the cryptocurrency boom, and the next era of micro-entrepreneurship is the most rammed 〈ChatGPT catches 18 fraud cases in New Taipei City! Civil servants of the audit department used APP to chase 2 million verdicts" This article was first published in BlockTempo's "Dynamic Trend - The Most Influential Blockchain News Media".