With the rapid development of the digital economy, artificial intelligence is undoubtedly one of the most innovative and promising fields today. Many AI entrepreneurs and programmers have enthusiastically entered this space, driven by passion and technological aspirations. However, seemingly innovative business models may also harbor numerous overlooked legal risks.
Through the first case in Shanghai involving illegal stock recommendations facilitated by an “AI stock-trading robot,” this article provides an in-depth analysis of the legal risks that AI entrepreneurs, programmers, and technical teams may face when engaging in fintech, quantitative trading, and related fields, and offers compliance recommendations.
I. Author: Attorney Shao Shiwei
1
Shanghai’s First Case of Illegal Business Operations Involving AI Stock-Trading Software: Company Controller Sentenced to Seven Years and Nine Months’ Imprisonment
Case Overview:
Without approval from the competent authorities, Company S operated the “Xundong Quant” online platform to promote stock-trading products such as Company S’s “range arbitrage” and “DIY stock-trading robot,” and provided securities consulting services to clients, including timing recommendations for buying and selling individual stocks and specific stock recommendations. The total illicit gains amounted to more than RMB 30 million.[i]
After adjudication by courts at two levels, Zhong, the actual controller of Company S, was ultimately sentenced to seven years and nine months’ imprisonment.
2
Is it unlawful to sell AI-based quantitative trading software for stock trading?
In this case, Zhong, the owner of the technology company, believed:[ii]
The DIY stock-trading robot he developed helped clients screen and purchase desired stocks based on plans set by the clients themselves. In Zhong’s view, all trading strategies were formulated by the clients, and the software merely performed data analysis, without constituting stock recommendation activities; therefore, the company’s operations did not require corresponding qualifications.
However, this view was not accepted by the court. Many practitioners in the AI industry, including programmers, may hold similar views:
AI-powered stock trading software serves merely as an information screening tool, not as legal advice.The software helps users improve information processing efficiency by automatically scraping and organizing publicly available market data (such as capital flows and trading volumes). The ultimate decision-making authority rests entirely with the user, which is fundamentally different from "stock recommendation" services that provide specific buy/sell advice or promise returns. Therefore, they argue that such software constitutes "neutral information technology services" rather than "illegal securities investment advisory business," and should not constitute the illegal operation of securities business as stipulated in Article 225 of the Criminal Law.
In this view, "factors" (such as unusual capital movements or sector anomalies) and "interval arbitrage" models,perform calculations and present results based on public, objective dataWhat the software does is merely simplify the information processing process, without making subjective "value judgments" or "investment decisions." Users trade according to their own settings, and the software automatically executes only when conditions are triggered. This constitutes instrumental execution following the user's autonomous decision-making, rather than the company making decisions on behalf of the user.
If the mere use of data analysis and automation technologies were to constitute a crime, thenall AI software or financial data terminals on the market that provide similar information screening and quantitative tools(such as certain features of Tonghuashun and East Money)could also be deemed illegal operations, which is clearly unreasonable.
So, why was Company S's business deemed to constitute the crime of illegal business operations?
3
Why the Court Found It Constituted the Crime of Illegal Business Operations: A Breakdown of Company S's Business Model
Based on Company S’s business model, the actual operations went far beyond the mere sale of AI-powered stock-trading software. The company not only provided strategy services such as “range arbitrage,” but also tiered its stock-trading robot software into different membership levels, namely VIP memberships priced at RMB 8,800 and RMB 28,800. Through these membership services, Company S charged customers an “interface usage service fee,” which was effectively a charge for “illegally accessing brokers’ trading channels,” ultimately generating cumulative profits of more than RMB 3 million.
Specifically, members at the RMB 8,800 tier received services such as “range arbitrage.” The term “range arbitrage” refers to the software’s analysis of a stock’s price movements over several days to calculate reference values, recommend timing for buying and selling specific stocks to customers, or directly provide trading strategies to assist customers in quantitative investment. If customers were dissatisfied with the software’s preset algorithms and trading strategies, they could obtain further analysis and recommendations from customer service personnel.
Members at the RMB 28,800 tier could, based on the data, models, and parameters provided by the company, set their own investment “tracks” and capital allocation plans. Once the preset conditions were triggered, the software would automatically execute buy and sell orders on behalf of the customers.
Therefore, the reasons why Company S constituted the crime of illegal business operations are as follows:
The essence of the RMB 8,800-tier service was to provide customers with specific investment advice—recommending timing for buying and selling individual stocks, providing strategy-based trading, and having customer service personnel offer analyses and recommendations based on experience. This service directly amounted to “telling customers what to buy and when to buy,” with its core being the provision of investment advice, which clearly exceeded the scope of an information intermediary.
The RMB 28,800-tier service, by combining the data and models provided by the company, directly executed trades on behalf of customers. The software’s automatic execution function essentially assumed the role of a trading channel and order execution mechanism, which lies at the core of a securities broker’s brokerage business.
The combination of these two aspects enabled Company S to bypass the role of licensed securities firms and directly carry out core securities trading and investment advisory businesses that should have been conducted by licensed institutions such as securities companies and investment consulting firms.
4
Compliance Boundaries for AI-Powered Stock-Trading Software: Criteria for Distinguishing Criminal from Non-Criminal Conduct
In light of the above case, does this mean that it is legally prohibited for relevant individuals or companies to sell AI-powered stock-trading software without obtaining qualifications and licenses such as the “License for Operating Securities and Futures Business” or the “Qualification for Sales of Securities Investment Funds”? Certainly not.
In practice, the core of the controversy does not lie in whether AI was used or whether quantitative methods were employed, but rather in whether the final output of the products and services consists of data and tools, or “investment advice and trade execution.” In other words, the question is whether the provider has moved beyond neutral technical services and crossed the boundary into securities investment advisory or securities business operations.
In light of Company S’s business model, a comparative explanation can be provided from the following three perspectives:
1. Functional Output Perspective:
Within Company S’s services, the “range arbitrage” model and customer service recommendations output explicit instructions on “when to buy or sell which stock.” This completes the transition from “what it is” (data facts) to “how to do it” (investment decisions), which is precisely the essence of investment advisory business.
According to Article 1 of the Interim Provisions on Strengthening the Supervision of Securities Investment Advisory Business Conducted by Using “Stock Recommendation Software” (Second Revision in 2020): Software products, software tools, or terminal devices that possess functions for aggregating securities information or compiling historical data statistics on securities investment varieties, but do not have functions such as “providing investment analysis opinions on specific securities investment varieties, predicting price trends, offering variety selection suggestions, or providing actual buying and selling recommendations,” do not constitute “stock recommendation software.”
Therefore, if an AI tool only provides objective data displays such as securities information aggregation, historical data statistics, capital flow, or sector anomalies, and does not provide analysis opinions, trend predictions, variety selection suggestions, or trading recommendations for specific securities, it is closer to a neutral information tool.
2. From the perspective of business logic:
Company S’s business model involves charging high, tiered membership service fees. The basis for these charges is the “provision of investment advice and trading conveniences capable of generating profits.” From the user’s perspective, they are paying for “who can quickly execute profit-generating operations on my behalf.” This is essentially the sale of investment advisory services and trading channel services.
In contrast, if a product is sold only as a one-time software purchase or charges via data subscription fees, with its primary value lying in the information tool itself and not using “profitability” as a core selling point, it typically aligns more closely with the business logic of technical services or information services.
3. From the perspective of operational closure:
Within Company S’s services, particularly in the RMB 28,800 tier, after users set conditions, the software automatically completes the entire process from decision triggering to order execution. This has partially replaced the roles of investors and securities brokers.
In contrast, if a product only provides analysis tools, requiring users to make their own judgments and place their own orders, with the platform not substituting for the decision-making and execution stages, its overall risk boundary is relatively controllable.
5
The addition of “black technology” leads to combined punishment for multiple crimes
The reason Company S’s software can rapidly assist clients in automated stock trading is that it purchased plug-in programs written by Han, a “hacker” who developed such external programs, integrated them into its company programs, and deployed them on company servers. This enabled Company S’s accounts to bypass the relevant technical protection measures of Tongdaxin Software (the market data terminal provider), invoke its trading channels, illegally access securities company servers, and conduct automated stock trading, thereby charging clients interface usage service fees.
In this regard, Han was sentenced to three years in prison for the crime of providing programs for intruding into computer information systems. Zhong, the actual controller of Company S, and Kong, the company’s technical director, were both subject to combined punishment for multiple crimes, namely illegal business operations and copyright infringement.
The reason is that, under the compliance framework, the basic structure of securities trading should be as follows:
(I still prefer the most primitive method of drawing diagrams)
The stock exchange is responsible for trading rules and matching mechanisms; securities companies, as the only lawful operating entities, establish client relationships with investors, and investors participate in securities trading through securities companies. Market data and trading terminal service providers such as Tongxinda provide technical tool support to securities companies and investors; they themselves neither engage in the operation of securities business nor establish securities service relationships with investors.
However, in this case, due to the involvement of Company S, the structure evolved into the following:
Investors first entered into transactional relationships with Company S by purchasing so-called "range arbitrage" and "stock-trading robot" products and services. Company S then illegally accessed brokers' trading channels by cracking Tongxinda's interface programs and bypassing normal technical protection measures, thereby completing automated trading operations and continuously charging investors for profit.
Under this structure, the brokers and Tongxinda were not the organizers or beneficiaries of the business, but rather passively became "objects of exploitation" within its technical pathway. The entity that truly constructed the business model, controlled the trading pathway, formed a closed-loop fee collection system, and profited therefrom was always Company S itself.
6
Risk warnings for AI entrepreneurs and programmers:
This case, in which a technology company received criminal sentences for its "AI stock-trading software," essentially exposes the ambiguous compliance boundaries surrounding numerous AI startup projects, quantitative trading products, and fintech tools.
For AI entrepreneurs, company executives, technical co-founders, programmers, and even product managers, if you are engaged in or planning to enter related fields, it is advisable to conduct a prudent assessment of your products and business models to ensure they do not cross the regulatory boundaries of securities business, thereby avoiding being mischaracterized as providing investment advice or engaging in disguised securities business operations.
[i] Case No.: (2024) Hu 0106 Xing Chu 836, et al.
[ii] Developer of "DIY Stock-Trading Robot" Sentenced! https://mp.weixin.qq.com/s/pC4GKGwL94rAJP1Yf7GXUg
Recommended Reading
