Transcript | Seeking the Next DeepSeek: the Chinese Generative AI Algorithm Registry, with Kendra Schaefer

Below is a complete transcript of the episode. Thanks to CadreScripts for their great work, to Oana Grigor and Natalia Polom for checking and formatting, and to Zhou Keya for the image! Listen in the embedded player above.

Kaiser Kuo: Welcome to the Sinica Podcast, a weekly discussion of current affairs in China. In this program, we’ll look at books, ideas, new research, intellectual currents, and cultural trends that can help us better understand what’s happening in China’s politics, foreign relations, economics, and society. Join me each week for in-depth conversations that shed more light and bring less heat to how we think and talk about China.

I’m Kaiser Kuo, coming to you from my home in Chapel Hill, North Carolina.

Sinica is supported this year by the Centre for East Asian Studies at the University of Wisconsin-Madison, a national resource centre for the study of East Asia. The Sinica Podcast will remain free, but if you work for an organisation that believes in what I’m doing with the podcast, please consider lending your support. You can get me at sinicapod@gmail.com. And listeners, please, please, please support my work at sinicapodcast.com. Do become a subscriber and enjoy, in addition to the podcast, the complete transcript of the show, absolutely invaluable, essays from me, as well as writings and podcasts from some of your favourite China-focused columnists and commentators. Do check out the page to see all that’s on offer.

And again, consider helping me out. Also, check out the new show — China Talking Points — now available on YouTube, and streaming live every other Wednesday. This week we’re going to be back. It’s going to be, again, the three of us: Andrew Polk and Eric Olander and myself. This week on Sinica, I am thrilled to welcome back one of the smartest, clearest-eyed analysts of tech in China, Kendra Schaefer, Head of Tech Policy Research at Trivium China.

Those of you who follow the show know that Sinica has been partnering quite closely with the team at Trivium, and today we get to showcase one of their standout pieces of recent work from one of their standout analysts. In January of this year, an AI startup called DeepSeek, of course, stunned the world by dropping a highly capable large language model that rivaled top-tier Western offerings. That debut raised a lot of eyebrows, and not a few questions about where China’s generative AI ecosystem really stands. Enter Kendra’s report, Seeking the Next DeepSeek. It digs into a remarkable data set maintained by China’s cyberspace regulator, the CAC, which requires registration of all public-facing generative algorithms. She told me about the database and this work she was doing on it over coffee last month when I was visiting Pittsburgh.

That was great. And a couple of weeks ago, she put it out. The report and the database offer a rare window into what’s actually being built in China, built and deployed, not just by the giants like Tencent, Alibaba, ByteDance and Baidu, but also by hundreds of startups, state labs, and even government agencies. As of April 2025, this list included 3,739 registered generative algorithmic tools, or GATs, that’s an acronym you’re going to be hearing a bit, ranging from foundational LLMs like DeepSeek to B2C voice assistants, image generators, video generators. Kendra’s analysis gives us the most comprehensive look to date at where generative AI innovation is happening in China — who’s doing it, which sectors are really heating up, where foundational labs are proliferating, and what role the state is playing in shaping this landscape?

It also offers a window into China’s evolving innovation strategy, where it’s succeeding, where it’s duplicating efforts, and where the next breakthroughs might come from. And the report illustrates, really, the value of mining China’s regulatory disclosures when you know where to look. There’s actually a surprising amount of actionable and public info on AI development in China. So, let’s get into it with Kendra Schaefer. Kendra, welcome back to Sinica.

Kendra Schaefer: Thank you. So good to be here as always.

Kaiser: Well, before we get into the data set itself, there’s the very reason it exists in the first place. And that is, of course, that it was mandatory to register these generative AI algorithms, as I mentioned just now. Folks interested in the AI regs and on their genesis can listen to the episode that you did on Sinica with Jeremy Daum, which was great. But for today, Kendra, can you give us a quick overview on where else in the world this type of regulation currently exists? Do we have something like this in the U.S.? Does the EU have something like this?

Kendra: No. And that’s why this data set is so interesting. And it’s also one of the reasons this data set is relatively unknown. I mean, where else in the world can you get a complete list of all of the different generative AI tools operating within the borders of a country? Now, there are some caveats about what’s actually on the list and what’s actually not on the list. But this is incredibly unique. It’s not just unique for the era. It’s unique globally. The U.S. does not require generative AI tools to register with the state, obviously. And Europe, even though they do regulate algorithms quite heavily, they haven’t, certainly not this heavily, and they haven’t taken this tactic, where they require companies to kind of file with a regulator for any tool that is interacting with the general public.