AI Regulation Tracker / Enforcement and orders
China SAMR Publishes Five AI-Related Unfair-Competition Cases
On February 6, 2026, China's State Administration for Market Regulation (SAMR) published five typical cases of unfair competition in the AI field. All five are concluded enforcement actions with penalties already imposed, ranging from RMB 5,000 to RMB 360,000. Two involved firms trading on the "DeepSeek" name, one on "ChatGPT," one an AI voice-dialing scam, and one a trade-secret unfair-competition case involving AI research and development, brought under the Anti-Unfair Competition Law. This is an enforcement summary under existing law, not new AI rulemaking.
SAMR framed the release against a policy backdrop. It cited the Fourth Plenum of the 20th Central Committee, which called for fully implementing an "AI Plus" action and for strengthening AI governance. Against that, the regulator said it is both protecting AI-industry development from copycats and cleaning up bad conduct that uses AI as a tool. In its words, the goal is to "lawfully handle unfair competition conducted through AI technology, guiding technology toward the good." The five cases are the evidence it chose to publish.
What the five cases actually punished
The pattern matters more than any single fine, so here is each in plain terms. The first case, handled by the Chaoyang district bureau in Beijing, hit a company that promoted a "DeepSeek local deployment tool" and plastered the "DeepSeek" name and official-style logo across its site, then bought paid search ranking to ride DeepSeek's fame, despite having no connection to DeepSeek. The fine was RMB 5,000.
The second, from the Xuhui district bureau in Shanghai, targeted a WeChat public account called "ChatGPT Online" that actually called OpenAI's public API, presented itself as a "Chinese ChatGPT," and used a logo highly similar to the official one. The regulator treated that as commercial confusion and fined the operator RMB 62,692.7.
The third case is the one I would flag hardest for anyone in financial services. A Shanghai company sold and ran an AI voice auto-dialing product that used preset scripts to place calls and screen prospects. It helped loan intermediaries place calls that invoked bank names or "free loan agency" scripts, knowing those intermediaries were neither banks nor bank partners. SAMR called that assisting false advertising and fined the company RMB 200,000.
The fourth case moved from marketing into trade secrets. An engineer at a computing company, bound by a confidentiality agreement and working on AI big-data product development, used his work key to log into the company test server from home and download 15.88 GB of files containing complete code including data algorithms. The material was found to be a protected trade secret, and the penalty was RMB 360,000, the largest of the five. The fifth case, from the Xihu district bureau in Hangzhou, was another "DeepSeek local deployment" website that copied DeepSeek's marks and induced users to pay, drawing a RMB 30,000 fine.
Why the legal basis, not the technology, is the story
None of these five turned on a dedicated AI law. The regulator reached for the Anti-Unfair Competition Law, as amended in 2019, and the Interim Provisions on Online Unfair Competition. Cases one and five were confusion and passing-off under Article 6. Case two was also confusion. Case three was assisting false and misleading advertising. Case four was trade-secret misappropriation under Article 9. The AI element is what made the conduct scalable and newsworthy, but the liability hooks are ordinary market-conduct rules that have existed for years.
That is the useful signal. A regulator does not need a bespoke AI statute to act. It can treat AI branding, AI marketing, and AI model code as the same categories of protectable interest it already polices, and it will publish the results to warn everyone else. The DeepSeek and ChatGPT examples also show that a famous AI name is now a magnet for copycats, and the regulator will move to protect the genuine brand even where the imitator adds a thin technical wrapper such as reselling an API.
Why this reaches US practice
SAMR does not regulate anyone in the United States, and this release binds only the five named parties. It reaches US professionals in two ways. First, if your company runs AI products or AI-branded services in the Chinese market, these fact patterns are a checklist of what gets penalized: copying a rival's AI brand, dressing an API reseller as the original, using AI dialing or generation tools to push deceptive claims, and letting staff walk off with model code. Second, if you own a well-known AI brand, this shows a major regulator willing to act against imitators trading on your name, which is a point worth knowing when you plan enforcement abroad. The discipline for US counsel and operators is the same one that already governs advertising and trade secrets at home. Confirm your marketing is accurate, your brand use is authorized, and your algorithms are access-controlled, because a capable regulator will read AI conduct through those familiar rules.
Questions professionals are asking
Is this a new AI law in China?
No. It is a compilation of five concluded enforcement cases, published together on February 6, 2026. Each case was decided under existing law, chiefly the Anti-Unfair Competition Law as amended in 2019 and the Interim Provisions on Online Unfair Competition. The release is an enforcement and guidance signal, not new rulemaking.
What were the fines?
They ranged from RMB 5,000 to RMB 360,000. The largest, RMB 360,000, went to a trade-secret case in which an engineer downloaded 15.88 GB of algorithm and code files. The others were RMB 5,000 and RMB 30,000 for two DeepSeek copycat sites, RMB 62,692.7 for a "Chinese ChatGPT" confusion case, and RMB 200,000 for assisting false advertising through an AI voice-dialing product.
Does this affect US companies?
Not as binding law. It matters if you operate AI products or AI-branded services in the Chinese market, because it maps the conduct that gets penalized, and it matters if you own a well-known AI brand that copycats might exploit there. For US readers it is a comparative enforcement benchmark applied through ordinary unfair-competition and trade-secret rules.
Why did brands like DeepSeek and ChatGPT appear?
Two cases involved firms using the "DeepSeek" name and logo on unrelated "local deployment" tools, and one involved a "ChatGPT Online" account that resold OpenAI's API while posing as an official Chinese version. The regulator treated these as commercial confusion and passing-off, protecting the genuine brands against imitators trading on their fame.
RELATED BRIEFINGS
Browse the full AI Regulation News tracker
Informational analysis for working professionals, not legal advice. Confirm how any decision, statute, or requirement applies to your situation with qualified professionals in the relevant jurisdiction.