China’s SAMR Releases Typical Cases of Unfair Competition Involving AI
Published 13 February 2026
Xia Yu
On 6 February 2026, the State Administration for Market Regulation of China (“SAMR”) released five typical cases concerning unfair competition in the artificial intelligence sector (“Typical Cases”), delineating the boundaries of fair competition in the commercial application of AI technologies. For AI enterprises, these Typical Cases not only reveal the current enforcement priorities of Chinese regulators, but also provide clear behavioral guidance in areas such as brand identifier use, advertising compliance, and protection of core technical secrets.
Overview of Typical Cases
Typical Case 1: Beijing Aoland Information Technology Co., Ltd. – Online Confusion
The party concerned promoted “DeepSeek Local Deployment Tool” software through its website, using the “DeepSeek” word mark and official icon without authorization, and exploited paid keyword bidding technology to free-ride on DeepSeek’s goodwill, causing users to mistakenly believe that it was affiliated with DeepSeek’s official operations. The Beijing Chaoyang District Market Supervision Administration held that this conduct violated Article 6 (3) of the Anti-Unfair Competition Law of the People’s Republic of China (2019) (“Anti-Unfair Competition Law”) and Article 7 (1) (iii) and (vi) of the Interim Provisions on Regulating Online Unfair Competition [ https://www.gov.cn/zhengce/202406/content_6959668.htm ], constituting the unauthorized use of another party’s well-known commercial identifiers to commit online confusion. The party was ordered to cease the infringing conduct and was fined RMB 5,000 (approximately US$ 700).
Typical Case 2: Shanghai Shangyun Network Technology Co., Ltd. – Online Confusion
The party concerned developed and operated a WeChat public account named “ChatGPT Online”, providing AI dialogue functions and generating revenue through registered paying members. The account used an avatar highly similar to OpenAI’s official “ChatGPT” icon and described itself in its profile as the “ChatGPT Chinese Version”. Investigation revealed that the party actually provided basic model services by invoking OpenAI’s application programming interface (API); it was neither the “ChatGPT” product itself nor affiliated with OpenAI in any way. The Shanghai Xuhui District Market Supervision Administration held that this conduct violated Article 6 (4) of the Anti-Unfair Competition Law, constituting the unauthorized use of another party’s well-known product names or identifiers in a manner sufficient to cause misidentification of a specific affiliation. In view of the party’s voluntary deletion of infringing content and cooperation with the investigation, the authority, after comprehensive discretion, ordered cessation of the violation and imposed a fine of RMB 62,692.70 (approximately US$ 8,777).
Typical Case 3: Shanghai Qiaopin Network Information Technology Co., Ltd. – Aiding False Advertising
The party concerned was primarily engaged in the sale, operation and maintenance of “Douxing Intelligent” robotic voice AI outbound call software. The software could automatically place calls using pre-set script templates, analyze call content, and screen potential customers for users. Investigation revealed that the party, despite being fully aware that its clients (various loan intermediary companies) were not banking institutions, had no cooperative relationship with banks, and did not provide fee‑free loan agency services, nevertheless used the said software to assist its clients in placing calls using scripts that impersonated banks or falsely claimed to offer “free loan agency services” to promote loan intermediary businesses. The Shanghai Market Supervision Administration held that this conduct violated Article 8 (2) of the Anti-Unfair Competition Law, constituting aiding other business operators in engaging in false or misleading commercial publicity. The party was ordered to cease the violation and was fined RMB 200,000 (approximately US$ 28,000).
Typical Case 4: Min XX – Misappropriation of Algorithmic Trade Secrets
Min XX, a development engineer at a computing company bound by a confidentiality obligation and responsible for R&D in AI big data products, on 21 August 2024, used his personal computer at his residence, logged into the company’s test server with his work credentials, and was found to have unauthorizedly downloaded a total of 15.88 GB of complete code files containing proprietary data algorithm technologies. Upon appraisal, the technical information involved was determined to constitute trade secrets. The Hangzhou Market Supervision Administration, Zhejiang Province, held that this conduct violated Article 9 (1) (i) and (2) of the Anti-Unfair Competition Law, constituting acquisition of a right holder’s trade secrets by theft, electronic intrusion, or other improper means. The party was ordered to cease the violation and was fined RMB 360,000 (approximately US$ 50,000).
Typical Case 5: Hangzhou Boheng Culture Media Co., Ltd. – Online Confusion
The party concerned established a website named “DeepSeek Local Deployment” to induce users to pay for services. Without authorization, the website used the identical “DeepSeek” logotype and highly similar icons, and its overall page design, the wording “DeepSeek Local Deployment Tool”, icons, and other content bore substantial similarity to the official DeepSeek website, sufficient to cause users to misperceive a specific affiliation between the website and DeepSeek’s official operations. The Hangzhou Xihu District Market Supervision Administration held that this conduct violated Article 7 (1) (iii) of the Interim Provisions on Regulating Online Unfair Competition and Article 6 (3) of the Anti-Unfair Competition Law, constituting the unauthorized use of another party’s well-known website names, page designs, or other identifiers to commit online confusion. The party was immediately ordered to cease the violation and was fined RMB 30,000 (approximately US$ 4,200).
Analysis of Typical Cases
The five Typical Cases identified above illustrate three prevalent patterns of unfair competition in the AI sector: online confusion (Typical Cases 1, 2, and 5), aiding false advertising using AI technology (Typical Case 3), and misappropriation of algorithmic trade secrets (Typical Case 4).
In Typical Cases 1, 2, and 5, the parties concerned each free‑rode on another party’s commercially well‑known AI product names, identifiers, or page designs, causing users to mistakenly believe that an authorization, cooperation, or other specific affiliation existed with the right holder. The Interim Provisions on Regulating Online Unfair Competition, promulgated by SAMR in 2024, expanded the scope of conduct constituting online confusion. Article 7 of it provides an expansive interpretation of Article 6 of the Anti-Unfair Competition Law in the digital context. The Article explicitly extends the protection against imitation and confusion to the following online identifiers:
1. Names, icons, and shapes of application software, mini‑programs, online stores, client terminals, public accounts, and game interfaces.
2. Online appellations, symbols, and abbreviations.
3. Explicit use of search keywords (i.e., setting another party's commercial identifiers as keywords for one's own paid keyword ranking).
This means that the protection enjoyed by AI enterprises for their commercial identifiers is no longer limited to registered trademarks or registered enterprise names but extends to unregistered digital brand assets that have “a certain degree of influence”, including product names, UI designs, and social media account avatars. Concurrently, the threshold for establishing infringement has been correspondingly lowered — the right holder need not prove the “likelihood of confusion” required for trademark infringement; it suffices to demonstrate that the conduct is “sufficient to cause misidentification of a specific affiliation”, thereby triggering administrative enforcement.
Typical Case 3 rejected the “technological neutrality” defense in the context of aiding false advertising. The party concerned was not the direct perpetrator of the false advertising, but rather the developer and seller of the AI outbound call software. Its core defense logic typically rested on “technological neutrality” — the software itself could be used lawfully, and the unlawful scripts were imported by the clients themselves. However, the enforcement authority did not accept this defense. Its reasoning can be summarized as follows:
1. Knowledge: The party had long served the loan intermediary industry and had clear knowledge of the facts that its clients were not banking institutions, had no cooperative relationship with banks, and did not provide fee‑free loan agency services.
2. Substantive Participation: The scripts in question were pre‑set or customized by the party concerned, not independently generated by the clients. The software automatically placed calls and analyzed call content based on these scripts; the party was thus deeply involved in the execution of the false publicity.
3. Pecuniary Interest: The party charged its clients based on call duration or closed transaction volume, creating a direct economic linkage between its financial interests and the clients' false publicity activities.
The implication of this case for AI enterprises is that AI technology service providers cannot evade liability on the grounds of “technological neutrality”. Where a service model involves the pre‑setting, recommendation, or optimization of user content, or where fees are calculated based on marketing performance, the provider may be found jointly liable as a facilitator of unfair competition.
Typical Case 4 applied the “acquisition as infringement” rule to a case involving algorithmic trade secrets and stands as a landmark case in administrative enforcement of trade secret protection in the AI sector in recent years. Its core adjudicative logic can be summarized as follows:
1. The act of downloading itself constitutes “acquisition by improper means”: Under Article 9 (1) (i) of the Anti-Unfair Competition Law, “theft, bribery, fraud, coercion, electronic intrusion, or other improper means” are typical forms of acquisition‑based infringement. The Interpretation on Several Issues Concerning the Application of Law in Handling Criminal Intellectual Property Cases, issued by the Supreme People’s Court and the Supreme People’s Procuratorate in 2025, further clarifies: “illegal reproduction” constitutes theft; “accessing an information system and acquiring trade secrets without authorization or beyond authorization” constitutes electronic intrusion. In this case, although Min XX held legitimate work credentials, the scale of downloading (15.88 GB), the device used (personal computer), the location (residence), and the timing (non‑working hours) clearly exceeded the scope of normal performance of duties, constituting “access beyond authorization” amounting to electronic intrusion.
2. Establishment of trade secret infringement does not require “disclosure” or “use”: In this case, there was no evidence that Min XX had disclosed the downloaded code to any third party or used it to develop competing products. Nevertheless, the administrative penalty decision affirmed the finding of infringement. This reflects a fundamental principle of trade secret protection: “acquisition” by improper means is itself an independent form of infringement, and its unlawfulness does not depend on subsequent conduct.
3. The standard of proof for reasonable confidentiality measures is becoming pragmatic: The company involved had implemented key‑based access control for its test servers and had executed confidentiality agreements with its employees. The enforcement authority did not demand that the enterprise adopt technically impenetrable barriers, but rather assessed whether the confidentiality measures were commensurate with the management context. This approach offers an important reminder for algorithm‑intensive enterprises: the focus should not be on theoretical impregnability, but on consistency in policy implementation — including regular access rights reviews, timely deactivation of accounts upon departure, and retention of operation logs.
Comment
Although the five Typical Cases released by the SAMR involve relatively modest monetary sanctions, the institutional signals they convey should not be underestimated. They signify that China’s competition enforcement authorities have evolved from “observers” of the AI industry’s development to its “regulators”. Competition in China’s artificial intelligence sector has entered a new phase characterized by defined boundaries, clear red lines, and enforceable rules.
Overview of Typical Cases
Typical Case 1: Beijing Aoland Information Technology Co., Ltd. – Online Confusion
The party concerned promoted “DeepSeek Local Deployment Tool” software through its website, using the “DeepSeek” word mark and official icon without authorization, and exploited paid keyword bidding technology to free-ride on DeepSeek’s goodwill, causing users to mistakenly believe that it was affiliated with DeepSeek’s official operations. The Beijing Chaoyang District Market Supervision Administration held that this conduct violated Article 6 (3) of the Anti-Unfair Competition Law of the People’s Republic of China (2019) (“Anti-Unfair Competition Law”) and Article 7 (1) (iii) and (vi) of the Interim Provisions on Regulating Online Unfair Competition [ https://www.gov.cn/zhengce/202406/content_6959668.htm ], constituting the unauthorized use of another party’s well-known commercial identifiers to commit online confusion. The party was ordered to cease the infringing conduct and was fined RMB 5,000 (approximately US$ 700).
Typical Case 2: Shanghai Shangyun Network Technology Co., Ltd. – Online Confusion
The party concerned developed and operated a WeChat public account named “ChatGPT Online”, providing AI dialogue functions and generating revenue through registered paying members. The account used an avatar highly similar to OpenAI’s official “ChatGPT” icon and described itself in its profile as the “ChatGPT Chinese Version”. Investigation revealed that the party actually provided basic model services by invoking OpenAI’s application programming interface (API); it was neither the “ChatGPT” product itself nor affiliated with OpenAI in any way. The Shanghai Xuhui District Market Supervision Administration held that this conduct violated Article 6 (4) of the Anti-Unfair Competition Law, constituting the unauthorized use of another party’s well-known product names or identifiers in a manner sufficient to cause misidentification of a specific affiliation. In view of the party’s voluntary deletion of infringing content and cooperation with the investigation, the authority, after comprehensive discretion, ordered cessation of the violation and imposed a fine of RMB 62,692.70 (approximately US$ 8,777).
Typical Case 3: Shanghai Qiaopin Network Information Technology Co., Ltd. – Aiding False Advertising
The party concerned was primarily engaged in the sale, operation and maintenance of “Douxing Intelligent” robotic voice AI outbound call software. The software could automatically place calls using pre-set script templates, analyze call content, and screen potential customers for users. Investigation revealed that the party, despite being fully aware that its clients (various loan intermediary companies) were not banking institutions, had no cooperative relationship with banks, and did not provide fee‑free loan agency services, nevertheless used the said software to assist its clients in placing calls using scripts that impersonated banks or falsely claimed to offer “free loan agency services” to promote loan intermediary businesses. The Shanghai Market Supervision Administration held that this conduct violated Article 8 (2) of the Anti-Unfair Competition Law, constituting aiding other business operators in engaging in false or misleading commercial publicity. The party was ordered to cease the violation and was fined RMB 200,000 (approximately US$ 28,000).
Typical Case 4: Min XX – Misappropriation of Algorithmic Trade Secrets
Min XX, a development engineer at a computing company bound by a confidentiality obligation and responsible for R&D in AI big data products, on 21 August 2024, used his personal computer at his residence, logged into the company’s test server with his work credentials, and was found to have unauthorizedly downloaded a total of 15.88 GB of complete code files containing proprietary data algorithm technologies. Upon appraisal, the technical information involved was determined to constitute trade secrets. The Hangzhou Market Supervision Administration, Zhejiang Province, held that this conduct violated Article 9 (1) (i) and (2) of the Anti-Unfair Competition Law, constituting acquisition of a right holder’s trade secrets by theft, electronic intrusion, or other improper means. The party was ordered to cease the violation and was fined RMB 360,000 (approximately US$ 50,000).
Typical Case 5: Hangzhou Boheng Culture Media Co., Ltd. – Online Confusion
The party concerned established a website named “DeepSeek Local Deployment” to induce users to pay for services. Without authorization, the website used the identical “DeepSeek” logotype and highly similar icons, and its overall page design, the wording “DeepSeek Local Deployment Tool”, icons, and other content bore substantial similarity to the official DeepSeek website, sufficient to cause users to misperceive a specific affiliation between the website and DeepSeek’s official operations. The Hangzhou Xihu District Market Supervision Administration held that this conduct violated Article 7 (1) (iii) of the Interim Provisions on Regulating Online Unfair Competition and Article 6 (3) of the Anti-Unfair Competition Law, constituting the unauthorized use of another party’s well-known website names, page designs, or other identifiers to commit online confusion. The party was immediately ordered to cease the violation and was fined RMB 30,000 (approximately US$ 4,200).
Analysis of Typical Cases
The five Typical Cases identified above illustrate three prevalent patterns of unfair competition in the AI sector: online confusion (Typical Cases 1, 2, and 5), aiding false advertising using AI technology (Typical Case 3), and misappropriation of algorithmic trade secrets (Typical Case 4).
In Typical Cases 1, 2, and 5, the parties concerned each free‑rode on another party’s commercially well‑known AI product names, identifiers, or page designs, causing users to mistakenly believe that an authorization, cooperation, or other specific affiliation existed with the right holder. The Interim Provisions on Regulating Online Unfair Competition, promulgated by SAMR in 2024, expanded the scope of conduct constituting online confusion. Article 7 of it provides an expansive interpretation of Article 6 of the Anti-Unfair Competition Law in the digital context. The Article explicitly extends the protection against imitation and confusion to the following online identifiers:
1. Names, icons, and shapes of application software, mini‑programs, online stores, client terminals, public accounts, and game interfaces.
2. Online appellations, symbols, and abbreviations.
3. Explicit use of search keywords (i.e., setting another party's commercial identifiers as keywords for one's own paid keyword ranking).
This means that the protection enjoyed by AI enterprises for their commercial identifiers is no longer limited to registered trademarks or registered enterprise names but extends to unregistered digital brand assets that have “a certain degree of influence”, including product names, UI designs, and social media account avatars. Concurrently, the threshold for establishing infringement has been correspondingly lowered — the right holder need not prove the “likelihood of confusion” required for trademark infringement; it suffices to demonstrate that the conduct is “sufficient to cause misidentification of a specific affiliation”, thereby triggering administrative enforcement.
Typical Case 3 rejected the “technological neutrality” defense in the context of aiding false advertising. The party concerned was not the direct perpetrator of the false advertising, but rather the developer and seller of the AI outbound call software. Its core defense logic typically rested on “technological neutrality” — the software itself could be used lawfully, and the unlawful scripts were imported by the clients themselves. However, the enforcement authority did not accept this defense. Its reasoning can be summarized as follows:
1. Knowledge: The party had long served the loan intermediary industry and had clear knowledge of the facts that its clients were not banking institutions, had no cooperative relationship with banks, and did not provide fee‑free loan agency services.
2. Substantive Participation: The scripts in question were pre‑set or customized by the party concerned, not independently generated by the clients. The software automatically placed calls and analyzed call content based on these scripts; the party was thus deeply involved in the execution of the false publicity.
3. Pecuniary Interest: The party charged its clients based on call duration or closed transaction volume, creating a direct economic linkage between its financial interests and the clients' false publicity activities.
The implication of this case for AI enterprises is that AI technology service providers cannot evade liability on the grounds of “technological neutrality”. Where a service model involves the pre‑setting, recommendation, or optimization of user content, or where fees are calculated based on marketing performance, the provider may be found jointly liable as a facilitator of unfair competition.
Typical Case 4 applied the “acquisition as infringement” rule to a case involving algorithmic trade secrets and stands as a landmark case in administrative enforcement of trade secret protection in the AI sector in recent years. Its core adjudicative logic can be summarized as follows:
1. The act of downloading itself constitutes “acquisition by improper means”: Under Article 9 (1) (i) of the Anti-Unfair Competition Law, “theft, bribery, fraud, coercion, electronic intrusion, or other improper means” are typical forms of acquisition‑based infringement. The Interpretation on Several Issues Concerning the Application of Law in Handling Criminal Intellectual Property Cases, issued by the Supreme People’s Court and the Supreme People’s Procuratorate in 2025, further clarifies: “illegal reproduction” constitutes theft; “accessing an information system and acquiring trade secrets without authorization or beyond authorization” constitutes electronic intrusion. In this case, although Min XX held legitimate work credentials, the scale of downloading (15.88 GB), the device used (personal computer), the location (residence), and the timing (non‑working hours) clearly exceeded the scope of normal performance of duties, constituting “access beyond authorization” amounting to electronic intrusion.
2. Establishment of trade secret infringement does not require “disclosure” or “use”: In this case, there was no evidence that Min XX had disclosed the downloaded code to any third party or used it to develop competing products. Nevertheless, the administrative penalty decision affirmed the finding of infringement. This reflects a fundamental principle of trade secret protection: “acquisition” by improper means is itself an independent form of infringement, and its unlawfulness does not depend on subsequent conduct.
3. The standard of proof for reasonable confidentiality measures is becoming pragmatic: The company involved had implemented key‑based access control for its test servers and had executed confidentiality agreements with its employees. The enforcement authority did not demand that the enterprise adopt technically impenetrable barriers, but rather assessed whether the confidentiality measures were commensurate with the management context. This approach offers an important reminder for algorithm‑intensive enterprises: the focus should not be on theoretical impregnability, but on consistency in policy implementation — including regular access rights reviews, timely deactivation of accounts upon departure, and retention of operation logs.
Comment
Although the five Typical Cases released by the SAMR involve relatively modest monetary sanctions, the institutional signals they convey should not be underestimated. They signify that China’s competition enforcement authorities have evolved from “observers” of the AI industry’s development to its “regulators”. Competition in China’s artificial intelligence sector has entered a new phase characterized by defined boundaries, clear red lines, and enforceable rules.