"The 'sure profit' scam has been exposed."

Ask AI · How do high-yield AI scams exploit human greed to set traps?

On April 2, the three major A-share indexes opened lower and kept falling. By the close, the Shanghai Composite Index fell 0.74%, the Shenzhen Component fell 1.60%, and the ChiNext Index fell 2.31%. The total citywide trading volume was 1.86 trillion yuan, and more than 4,300 stocks declined.

What happens if you hand 100,000 yuan to AI to trade stocks?

Amid the craze for AI intelligent agents such as “Lobster,” AI stock trading has made many investors eager to give it a try. Wang Tao, who has 5 years of investment experience, is one of them. Half a month ago, he handed 100,000 yuan to AI in an attempt to test the returns from AI stock trading. Wang Tao believes that in the past, he was overly emotional: chasing rallies, cutting losses at the wrong time, and trading frequently, which led to investment failure. Can a rational AI be more reliable than himself? As of now, Wang Tao’s account has lost several thousand yuan, and his doubts about the uncertainty of AI stock trading are growing.

On social media platforms, many high-yield case studies are becoming content designed to attract traffic. Some overseas bloggers claim that a “Lobster” turning $50 capital into $2,980 within 48 hours, for a return rate of 5860%. Many domestic paid AI stock-trading courses use cases of this kind to promote the idea that AI is invincible and that investors can “enjoy the benefits without effort.” Some investors have been tricked into losses, suffering severe damages.

Is it really feasible to hand over your money entirely to AI for management? Zhang Feng, Deputy Director of the Shanghai Financial and Legal Research Institute, said that any tool claiming “AI can make money for you” is, in essence, highly worth being on guard against. Investment returns come from risk pricing and market game theory. AI itself does not create returns; it only optimizes information and efficiency. Promising to “make money on behalf of clients” is fundamentally misleading, and may even involve illegal securities recommendations and fraud.

AI illustration/adan

Worried about losing your job,

a programmer starts trading stocks with AI

Li Yanqi works as a programmer at an internet company in Beijing. His job is front-end testing. Over the past two years, the company has introduced an AI automated testing platform that can automatically generate test cases, identify visual deviations, and predict crash paths—more “skillful” than programmers. Since this year, he has found that AI coding speeds have been getting faster, and many basic programming tasks have already been replaced by AI. The company has also started reducing recruitment of new staff. His role has shifted from an executor to a supervisor, and it may very possibly become a “redundant position.”

Li Yanqi began to think about his future prospects and wanted to try “using AI to capture the dividends of the times and start trading stocks.” In the past, Li Yanqi also bought and sold stocks. His approach was no different from most retail investors: “After a few years, I didn’t make money; I even lost part of my principal.”

Li Yanqi decided to change his approach. Instead of manually monitoring the market himself, he began testing a model to replace him in making decisions. “AI doesn’t have human fears and greed, so it can be objectively rational and is not easily swayed by emotions. That’s a trait people lack in the stock market.” Leveraging his advantage as a programmer, he wrote a program to collect, organize, and compile everything: the day’s market trading data, the company’s latest news, and changes in industry policies—then send them to the model.

Li Yanqi wasn’t simply asking the model what stock to buy today. He gave the model several different roles, including acting as a macro analyst researcher, a risk controller, and an actual execution trader. The macro analyst is responsible for analyzing the overall situation of the market and identifying which industries are receiving policy support. The risk controller is responsible for assessing the probability of potential losses and reminding him when he needs to reduce the amount of capital invested. The trader is responsible for giving a specific buy-and-sell plan based on the opinions of the first two roles.

Li Yanqi asked the model to write an investment diary after the market closes every day, recording the discussion process of these roles. He found that the model’s investment philosophy is different from that of ordinary people. Ordinary people like to chase stocks whose prices rise quickly over a few days, while the model analyzes a company’s real ability to make money and recommends buying stocks whose prices are still at low levels and whose business has growth potential. The model will tell him that when facing price declines, you need patience—so long as the fundamentals of the company have not changed, you should hold long term, and you shouldn’t be scared off by the market’s short-term fluctuations.

In fact, on GitHub, open-source AI stock-trading tools frequently make the hot search list, and some projects have already gained thousands of stars. The AI stock-trading track is starting to come into the public’s view. Xing Xing, Chief Economist at Jin Donghui Enterprise Management Development (Beijing) Co., Ltd., told China News Weekly that A-share individual investors account for a high proportion, and their professional abilities vary widely; time and information disadvantages are obvious. There is strong demand for standardized, low-cost, disciplined asset allocation services. Combined with increased market volatility and a growing awareness of long-term allocation, AI happens to fill the supply gap of inclusive wealth management.

Xing Xing summarized that the benefits AI brings to investors mainly include: lowering information costs, quickly processing massive data, and reducing information asymmetry faced by retail investors; strengthening investment discipline by executing strategies through algorithms and overcoming human weaknesses such as chasing rallies and selling during panic; improving allocation efficiency by enabling personalized asset portfolios and dynamic rebalancing, which is suitable for ordinary investors to hold long term. At the same time, it provides continuous empowerment in risk warnings, position monitoring, and investor education and guidance, making professional wealth-management services more accessible to the general public.

Relying on large models may lead to even greater losses

Li Yanqi feels that AI models are like tireless assistants that can find information he usually overlooks. To verify the effectiveness of this approach, Li Yanqi took 20,000 yuan and placed it in a separate account for a live trading test. He bought and sold strictly according to the model’s recommendations and no longer trusted his own intuition. After running for a few months, he found that the account’s capital fluctuations became much smaller, and there was no longer the situation where he used to lose a lot of money in a single day.

He now has a bigger plan. He wants to turn this prompt-built method into an automatically running system, so that the program can fetch data, analyze it, and place orders on its own—achieving automated trading that operates independently of manual operations.

However, real investors have a different understanding of AI stock trading. Xie Minghui, a full-time instructor at the School of Economics and Management of Wuchang Institute of Technology, is a “trader” with more than 20 years of investment experience. His academic background in finance continuously shapes his understanding of live trading operations. Last year, his return rate was as high as 200%.

In his day-to-day trading, Xie Minghui has also started using AI. He believes that AI’s biggest current role is processing textual information. In the past, when researching a listed company, Xie Minghui needed to download hundreds of pages of annual financial reports from various websites, search page by page for key data, and also look up statements by company executives and changes in the industry—this took a lot of time. Now, after handing this information to AI, the model can help him extract key data such as changes in operating revenue, profit, and sales volumes of core products in a very short time, and organize it into a table. This approach saves an enormous amount of time.

Xie Minghui realized that there are many people around him who lack basic knowledge. When they hear that AI can trade stocks, they use it blindly without thinking. “These people don’t understand what the P/E ratio is, and they don’t understand value investing. They only know to let AI give them a stock code, then they put all their money in.” Xie Minghui believes that this approach has extremely serious hidden risks.

Taking corporate governance as an example, Xie Minghui found that in some listed companies, controlling shareholders hold a very large proportion of shares. When the stock price is driven up, these major shareholders will choose to sell the shares and cash out. “This kind of human change and calculation of interests is something that AI cannot predict from just a few financial statements. Retail investors only look at the technical chart analysis provided by AI and don’t understand the distribution of interests behind the company, so they can easily get trapped.”

At present, many tools on the market are called “intelligent investment advisors.” They provide users with various data analysis services. Xie Minghui believes these tools do have some advantages. They can break information asymmetry and can recommend different proportions of stock and bond portfolios based on the user’s risk tolerance. However, these tools also have obvious disadvantages. The results they generate all depend on data from the past and they don’t know what unexpected events will happen in the future.

“Right now, many financial platforms are rolling out related services, and some platforms have marketing practices that exaggerate returns and downplay risks, which easily causes short-term cognitive biases in investors.” Xing Xing also noticed this. AI models over-extract coincidental patterns from historical data and treat noise as signal—especially by doing targeted optimization for extreme returns such as limit-up situations. The more eye-catching the historical performance is, the more fragile the live-trading performance becomes. That’s why inherent data bias is formed.

He believes: “For investors, this could mean situations where AI cannot adapt in real time to regulatory requirements, to changes in capital behavior, or to changes in emotion. This means that many times, the investment advice provided by AI exists only in an ideal environment and cannot be replicated reliably in the real market.”

A new gimmick for harvesting investors

Some illegal actors have begun to use these “high-end” terms to create scams. Many investors who don’t understand technology become their target. On the Black Cat Complaints platform, there are rights-protection reports related to AI stock-trading software, and most of the victims are middle-aged and elderly people.

A netizen reported on the Black Cat Complaints platform that some savings in their family were held by elderly relatives, who had a need to preserve and increase the value of their money. A software company in Hebei named Yuan da took advantage of the elderly people’s mindset and developed a quantitative stock-trading software that claims to use the latest models to calculate buy and sell points, called “Fish Leaps Over the Dragon Gate.” When selling, salespeople showed the elderly people an image about making money. The image data showed that since the software started running in September 2024, the return rate reached 1000%.

Many elderly people bought the software for life at a high price, then followed the software’s prompts to buy and sell stocks. In the end, they didn’t make money; instead, their principal lost more than half. The netizen found that the software’s displayed “tenfold” return data had been manually modified in the backend, with the purpose of attracting investors to pay.

Even more outrageous are multiple “imposter” software companies represented by Wuhan Baiyu Quant and Shenzhen Yongjie Quant. According to an investigation by 21st Century Business Herald, these companies use “AI quantitative stock trading” as a publicity gimmick on the surface, but in reality they set up a fraudulent scam loop of “pyramid-style customer recruitment + Ponzi-like capital pool.” Under the banner of “AI quantitative trading” and “intelligent copying/following,” they claim to use their own developed AI systems to trade automatically, with a monthly return rate as high as 150%. In reality, the funds were never entered into securities accounts. Instead, they flowed into private accounts through third-party payment channels, and the profits investors see are merely numbers modified in the backend (a virtual account system).

The number of victims in these cases has reached several hundred, and the total amount of money defrauded has accumulated to tens of millions of yuan. Recently, many similar companies have experienced a concentrated collapse, and the police have already filed cases and launched investigations into some cases involved.

Yang Jianjun, a professor at the School of Law at Northwest University of Political Science and Law, told China News Weekly that currently there are three types of typical “AI guaranteed profit” scams, which mainly focus on AI training courses, AI stock trading and investment software, and AI automatic live-stream selling and content creation. These gimmicks often easily induce people to pay, and implement precise traffic redirection; in practice, multiple fraud cases have already emerged.

Yang Jianjun reminded investors that any project promising “low entry threshold, high returns, guaranteed profit” meets the basic characteristics of fraud. For software or services involving investment, investors must verify through official channels such as the China Securities Regulatory Commission whether the relevant institution holds legitimate qualifications. “AI is a powerful efficiency tool that can help us analyze information and generate content, but it is not a ‘magic touch for striking gold,’ and it cannot guarantee business success or investment returns. Don’t relax your vigilance just because the ‘AI’ label is put on it.”

A scholar in the finance field pointed out that currently, technological development is too fast, and regulation indeed has not fully caught up yet. On the one hand, as investors, people need to treat AI stock trading cautiously. On the other hand, government and regulatory authorities need to accelerate the drafting of relevant laws and regulations to prevent systemic risks from arising.

Yang Jianjun said that the industry is currently in intense discussion about whether general large models (such as DeepSeek) directly recommending stocks should be included in the regulatory scope of securities investment advisory licenses. With cases of investment losses caused by “AI hallucinations,” it is expected that within the next 2—3 years, mandatory provisions on information disclosure regarding AI investment advisory will be rolled out.

Earlier, the State Administration for Market Regulation and the National Standardization Administration issued the “Cybersecurity Technology — Safety Specifications for Data Labeling in Generative Artificial Intelligence.” The document specifies safety requirements for data labeling platforms or tools used for generative AI training, safety requirements for data labeling rules, requirements for data labeling personnel, requirements for data labeling verification, and describes data labeling safety evaluation methods.

From the perspective of legislative recommendations, Zhang Feng suggests that the four labeling requirements be incorporated as soon as possible into the revision of the “Administrative Measures for Securities Investment Advisory Business,” or into specialized financial AI regulatory regulations, to require strict implementation and strict law enforcement, and to promote compliant, transparent, and responsible development of AI in the financial sector.

(Wang Tao in the article is a pseudonym.)

Published in the April 6, 2026 issue, Total No. 1230, of China News Weekly

Magazine title: AI Stock Trading Chaos

Reporter: Meng Qian

Editor: Min Jie