When DeepSeek burst onto the global stage in January 2025, it seemed to appear out of nowhere. But the large language model was just one of the thousands of generative AI tools that have been released in China since 2023—and there’s a public archive of every single one of them.
The country’s top internet regulator, the Cyberspace Administration of China (CAC), requires that any company launching an AI tool with “public opinion properties or social mobilization capabilities” first file it in a public database: the algorithm registry. In a submission, developers must show how their products avoid 31 categories of risk, from age and gender discrimination to psychological harm to “violating core socialist values.”
Applicants submit their filing to their local CAC (say, the Shanghai CAC for Shanghai-registered firms), which forwards applications to the central CAC for final approval. Only then is a tool publicly listed in the algorithm registry. While the European Union is pursuing a single, comprehensive AI Act, notes Matt Sheehan, a research scholar at the Carnegie Endowment for International Peace, China’s approach to regulation is more ad hoc, targeting specific algorithms and building up iterative standards. (The US has no comparable registration system or centralized regulatory agency.)
Over time, the CAC has inadvertently created the most detailed map of a nation’s AI ecosystem anywhere in the world.
Open the CAC’s update from August 2024 and you’ll find DeepSeek listed as entry 152, a single row in a neatly packed table. Scroll through the table and you’ll find an AI that manages homestays and an AI that drafts patents. One assists ob-gyns in a Shanghai maternity ward; another helps manage state power grids. Kendra Schaefer and her colleagues at Trivium China, a Beijing-based policy consultancy, have been compiling the CAC’s updates into a comprehensive database, enriched with their own research.
A Broad View of the Boom
Nearly 80 percent of China’s generative AI registrations are clustered in and around its top tech hubs—Beijing, Shenzhen, Shanghai, and Hangzhou. Each city has its strengths: Beijing’s elite universities, national labs, and political strength give it an edge in large-scale innovation; Shenzhen (in Guangdong) is home to a dense hardware supply chain and vast pool of engineering talent; Shanghai, close to multinationals, excels at commercialization; and Hangzhou (in Zhejiang) is fueled by Alibaba’s ecommerce empire.
But innovation spreads far beyond the coasts. Chongqing is positioning itself as an AI manufacturing and logistics node; and heavy state investment has helped Hefei, in Anhui Province, become known as “China’s speech valley” for its cluster of speech-recognition firms, including iFlyTek. Filings also originate in less obvious regions like Guizhou, China’s “Big Data Valley,” where massive data centers power Huawei’s Pangu model, and Inner Mongolia, where state enterprises are integrating AI into mining and agriculture.
In the Trivium dataset, state-linked listings—from state-owned enterprises to government-backed research institutes—make up 22 percent of filings. Many state-linked firms partner with Big Tech to build their AI: PetroChina, for example, teamed up with Huawei and iFlyTek to create oil and gas applications; State Grid used DeepSeek to build a model optimizing power grids.
Foreign firms make up just 0.5 percent of filings. Ikea, for example, has a smart shopper algorithm that generates product recommendations. Yum China, the parent company that operates Kentucky Fried Chicken in China, listed a model that generates menus and promotional material.
Zeroing In on the Competition
More than half of the listings in the algorithm registry are for what Schaefer calls cross-sector technologies. These range from foundational models to “general purpose” text generators to a wide array of multimedia tools—voice swappers, 3D renderers, image makers. “Nobody wants to be caught in a situation where they depend on a competitor’s technology,” Schaefer says. Unlike in the US, where OpenAI, Anthropic, and Google DeepMind dominate the market, China’s competition to build foundational AI remains diverse and contested. But building these models is costly, and the market is beginning to consolidate. China’s six “AI tigers”—Moonshot, Minimax, Zhipu, Baichuan, 0.1AI, and Stepfun—are all backed by Alibaba or Tencent. ByteDance’s Doubao surpassed DeepSeek as China’s most popular chatbot, but its spot at the top is not assured.
Niche Natives
While the giants duke it out for chatbot supremacy, startups are hard at work in every sector imaginable.
Squirrel AI 松鼠
Founder Derek Li says his 12-year-old company is leaps beyond the ed-tech competition. They “put wheels on a horse,” he says, bolting AI onto their existing stale software. Squirrel claims to diagnose knowledge gaps, measure progress, and adjust lessons in real time.
When China banned for-profit tutoring in 2021, the company’s revenues collapsed overnight. It pivoted to licensing its platform to franchisees who also sold the company’s AI-powered tablets. Squirrel’s network includes more than 3,000 centers across China, serving 1.2 million students. Now, the company is eyeing expansion to the US.
Li, who withdrew his sons from a private school in Shanghai so that they could be home-schooled on Squirrel’s platform, says that “in the future, teachers won’t teach knowledge.” Instead, he says, “they’ll become data analysts, understanding learning reports and students’ ability, and psychologists, understanding emotions and shaping their personalities.”
AI Kanshe 看舌
AI Kanshe (translated as “AI Sees Tongue”) is a traditional Chinese medicine startup that analyzes health through images of the tongue, palms, and face. The company was founded by Li Wenhua, a former employee of Yaoshi Bang, one of China’s earliest online pharmaceutical platforms. A longtime student of tongue and hand diagnosis, Li wanted to combine the diagnostic methods of traditional Chinese medicine with modern machine vision. The company serves both consumers and health practitioners across clinics, pharmacies, and some hospitals, offering tools to support diagnosis and decisionmaking. Its model is trained on more than 100,000 annotated images of tongues, hands, and faces.
Zhongtan Puhui Cloud Technology 中碳普惠云科技
Founded in 2024 by Wu Song, a former Wall Street quant trader, Zhongtan Puhui Cloud Technology develops AI-driven tools for carbon accounting. The green transition, Wu says, still relies on cumbersome human labor that could be automated.
Zhongtan Puhui builds AI agents that handle a number of carbon accounting tasks, including carbon footprinting and emissions audits. Its clients range from China Minmetals Group and DHL to small and medium-sized exporters in the Yangtze River Delta.
Read the full article here
