Author: Lawyer Shao Shiwei
According to industry insider news, in May 2026, an operator of an AI gateway service publicly stated that they had been criminally detained for 37 days for illegally reverse-engineering, scraping, and reselling low-cost AI API resources, and is currently under a status of release on bail pending trial.
Although Lawyer Shao has also seen fellow lawyers commenting on this incident, stating that an AI gateway operator was criminally detained by Shanghai police, as there is no official announcement yet, the specific charges and handling of the case cannot be confirmed.
In the past two years, the demand for AI applications in China has exploded. However, due to regional restrictions on overseas large language models, more and more people are operating AI gateways. Simply put, domestic users want to use them but can't, so the gateway helps bridge the channel and charges a toll.
Therefore, using this case as an example, we can discuss: Can the currently booming business of AI gateways still be done? What risks might ordinary people face when operating an AI gateway?
The Lucrative Business of AI Gateways
According to a report by Xinhua News Agency, China's average daily Token usage has increased over a thousandfold from early 2024 to March 2026. The demand is clearly there.
However, if domestic users want to use overseas large models (OpenAI, Anthropic, Google, etc.), they encounter numerous hurdles such as network environment, payment channels, and identity verification.
Where there are hurdles, there is business. AI gateways have emerged accordingly. Currently, on many self-media platforms, it's claimed that AI API gateways are one of the most profitable projects in 2026, and this is not false.
AI gateways can also be understood as AI scalpers. They package interfaces from different AI model providers into a unified outlet, handling backend connections to all models for the user. Users also avoid the need for accessing restricted networks or figuring out foreign currency payments.
On platforms like Taobao, Xianyu, and Xiaohongshu, many such posts can be seen, with prices shockingly low.
So the question arises: With prices pushed so low, how exactly do the gateway operators make money?
Reselling free credits. Platforms like ChatGPT and Claude offer free credits upon registering new accounts. The bulk account merchants behind the gateway service register large quantities of accounts, exploit the platform's free credits, then use technical means to reverse-engineer these accounts' web interfaces into standard APIs, and sell them uniformly. The cost is almost zero.
Refund arbitrage. They batch-register official accounts, top them up, and call the APIs. What if the account gets banned? Apply for a refund. In most cases, the pre-loaded money can be recovered. This means they use your money to call the APIs, and if banned, they get the cost back from the official source, profiting from both ends.
Over-reporting Tokens. Official APIs strictly charge based on the number of Tokens used, but the gateway's billing system is written by the operator themselves. Normally, 1 Chinese character is about 1.5 to 2 Tokens. Some gateways adjust the multiplier in the backend, so your 1 Chinese character could be charged as 3 to 4 Tokens. Users have no way to verify this.
Model swapping. You pay for Claude Opus 4.7, but you might actually be calling a small open-source model. This is also why many users feel that models via gateways seem "dumbed down".
Data resale. They package and sell users' complete conversation records, especially high-quality training data like code in programming scenarios, reasoning processes, and engineering decisions, to model vendors. Why are gateways so cheap? They often profit by selling data.
This business has grown so large that even celebrities are entering the field. On May 1, 2026, Tron founder Justin Sun launched the AI gateway B.AI, with the slogan "One API Key = Claude + GPT + Gemini + Full Series of Domestic Large Models." On May 5, the cryptocurrency company WLFI, linked to the Trump family, launched WorldRouter, directly connecting AI calls with the cryptocurrency system.
But the greater the traffic, the greater the associated risks.
Why Could Running an AI Gateway Lead to Arrest?
Earlier, we outlined the profit models of AI gateways. As a gateway operator, one might be aware of the risks in such gray-area projects. However, after operating for some time and genuinely making money, and seeing peers doing the same without incident, one tends to become complacent.
From a legal perspective, the criminal risks of AI gateways mainly concentrate on three levels.
First, the business model itself may be illegal.
The computing resources of AI gateways are not obtained through legitimate procurement of API interfaces. Instead, they involve batch-registering accounts to exploit free credits or using technical means to reverse-engineer and obtain interface access. This no longer falls under the category of normal commercial agency.
Providing information relay and data processing services essentially constitutes a telecommunications value-added service. According to the "Regulations of the People's Republic of China on Telecommunications," operating such a business requires corresponding administrative permits. Operating without permission carries the risk of violating the crime of Illegal Business Operations.
Furthermore, overseas large model providers have access restrictions for users in China. Gateways help users circumvent these restrictions through proxy IPs, fabricated identities, etc., essentially assisting in bypassing the service provider's access conditions. If such acts are deemed as disrupting market order, they could also fall within the scope of the crime of Illegal Business Operations.
Second, lack of data security obligations.
AI gateways process a large volume of interaction data between users and models daily. User prompts, code snippets, business documents are all transmitted and processed through gateway servers. As the actual handler of the data, the gateway legally bears corresponding security management responsibilities.
However, the reality is that the vast majority of gateways have not established any data security management system—data storage locations, access control, and security measures are all non-existent. Once a data breach occurs, whether due to external attacks or internal mismanagement, the gateway, as a network service provider, could face criminal prosecution for the crime of Refusing to Perform Information Network Security Management Obligations. This crime targets precisely the act of "having a legal obligation but failing to fulfill it."
Third, illegal collection and sale of user data.
Some gateways package and sell user conversation logs to third parties, which is not an isolated phenomenon within the industry. However, the corresponding legal risk of this behavior is generally underestimated.
Conversation content between users and AI models often contains personal information, commercial secrets, and other sensitive data. Gateways rarely obtain explicit user consent when collecting this data, let alone inform users of the data's purpose and flow. Collecting and providing such information to third parties without consent, if serious, constitutes the crime of Infringing on Citizens' Personal Information.
The threshold for this crime is not high. According to relevant judicial interpretations, illegally obtaining, selling, or providing 50 or more pieces of personal information such as location tracking, communication content, credit information, property information, or 500 or more pieces of other personal information that may affect personal or property safety, such as accommodation information, communication records, health and physiological information, transaction information, meets the standard for prosecution. Considering the daily data processing volume of a gateway, reaching this threshold is not difficult.
Final Thoughts
Regarding the issue of AI gateways, Lawyer Shao does not wish to limit the discussion merely to whether operators will be prosecuted. The incident reflects issues the AI industry inevitably faces during its rapid development phase.
For users, gateways lower the usage barrier, but also expose users' sensitive data to an intermediate link without qualifications or security safeguards. Once problems occur, users might not even find a party against whom to assert their rights.
For model vendors, the existence of gateways consumes their technological investment and business models. Free credits are exploited in bulk, paid interfaces are reverse-engineered and misused, and pricing systems are undermined. Vendors are forced to divert significant resources from product development to risk control and countermeasures, and these costs ultimately get passed on to normal paying users. A deeper damage lies in the fact that when gateways dump computing power at low prices, the market's perception of the value of AI services is being distorted—users gradually come to believe that these capabilities should be almost free. This harms the sustainable development of the entire industry.
As a lawyer specializing in the new economy sector, Lawyer Shao has also been following the development of the AI industry and has served many practitioners in this field. How far an industry can go depends not on how fast it runs, but on whether it can establish a basic commercial order and foundation of trust. The AI industry is at a critical stage of transitioning from wild growth to standardized operation. The choices of every practitioner are shaping the future ecosystem of this industry.
A healthy AI industry requires vendors to continuously invest in technological R&D, users' data rights to be genuinely protected, and practitioners to participate in market competition in a compliant and responsible manner. These are prerequisites for the industry's long-term survival. What Lawyer Shao hopes to see is more practitioners choosing to do the difficult but right thing, making the foundation of this industry even more solid.
Special Disclaimer: This article is an original work by Lawyer Shao Shiwei. It represents only the author's personal views and does not constitute legal consultation or advice on specific matters.
