Greetings from a world where…

I’ll be presenting a co-authored working paper at an online webinar this Thursday, with the Berlin Contemporary China Network. Here’s a preview:

Emerging economies face significant challenges in managing safety risks from powerful technological systems. Indeed, many analysts have identified China as the most likely source of a major accident linked to emerging technologies. Yet, contrary to these expectations, China has achieved a remarkable safety record in certain technological domains, such as civil aviation and nuclear power. How? We theorize that, for industries in which one firm’s accident damages the reputation of all others, international industry associations can contribute to improved safety standards in emerging economies

…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).

Around the Horn (23rd episode)

It is about that time: let’s go back Around the Horn!

This is our 23rd episode. For new readers, here’s how it works (see ChinAI #333 for the previous edition):

• I give short previews of ten articles that caught my eye during a scan through my usual sources (all published within the past week or so). The title for each preview links to the original article in Chinese.

• Readers vote on next week’s feature translation by replying to the email and/or commenting on the post with the number of your preferred article. *I’ll give some added weight to votes from readers who are paid subscribers to ChinAI.

• The main idea is that any of these 10 links would have made for a great feature translation this week — like the recent Clipse album, there are no skips!

1) After my voice was “stolen” by AI, will I be replaced?

Summary: How are Chinese voice actors reacting to AI? One jumping off point for this story is voice actress Xueting Mu’s experience, in which she accused an Amazon series for cloning her voice for AI training and use.

Source: 极昼工作室 (Perpetual Light Studio), a reporting unit under Sohu that does longform human-interest stories.

2) HEYTEA falls off, DeepSeek Gets Defeated by this Company: Who Won the 2025 Battle of Best Brands?

Summary: NetEase DataBlog runs an annual survey of top Chinese brands, with this year’s iteration surveying 2024 respondents for 31 subcategories. One of the new categories this year: AI applications, where (ByteDance) Doubao topped DeepSeek.

Source: 网易数读 (NetEase DataBlog) — provides informative, data-backed analysis of a wide range of topics

3) AI Safety/Security Governance Research Report (2025)

Summary: Authored by four CAICT experts, this report provides an overview of: the AI governance field and core challenges, the current status of international cooperation, and case studies of application industries. From my quick scan, there were some interesting details on slides 12-13.

Source: 中国信通院 (CAICT) — a think tank under China’s Ministry of Industry and Information Technology.

4) China’s Venture Capital Market Emerges from Winter

Summary: In 2025, the total number of transactions in China’s VC market increased by 28%, but the total financing amount remained stable from the previous year. This readout concludes that, if 2023 was winter, then 2025 was early spring.

Source: IT桔子 (IT Juzi) — good source on financing and venture capital in technology fields.

5) The People who Resist AI

Summary: Through March 1, 2024, over 5,600 new books were launched on Tomato Novel (China’s top free web novel platform); across the same period in 2023, just 400 were launched. After suspecting one of her favorite authors of using AI-generated slop, Yiqiu Han posted on social media (looks like Xiaohongshu to me?) under the hashtag #反ai (#Anti-AI). To date, that hashtag has garnered 40,000 discussion threads.

Source: The Mirror [镜相工作室] — they seem to be an influential platform that analyze AI through a more critical lens (came across this WeChat public account for the first time through a Huxiu recommendation).

6) The first state-owned enterprise AI unicorn emerges!

Summary: China Telecom, one of China’s big three state-owned telecom operators, apparently has an AI subsidiary that just raised its initial round of capital from four state-backed strategic funds. This is a very different profile — what’s going on?

Source: 量子位 (QbitAI) — similar to Leiphone, a news portal that regularly covers AI issues. They’ve been publishing longer reports in recent years.

7) ChatGPT, is it trying to “cross the river by feeling for the stones” of Ant Afu

Summary: Chinese fintech giant Ant’s AI health management app (Ant Afu) has surpassed 10 million daily consultations; now ChatGPT has launched a health function. Is American AI learning from Chinese AI in this space?

Source: 数字力场 (Shuzili chang) — a WeChat account that I’ve followed since 2021, but I haven’t really featured much of their work. One of their past articles was authored Zongming Yu, a former deputy editor of the commentary section in The Beijing News.

8) AI Verbally Abuses User, Tencent Yuanbao Responds

Summary: A netizen recently shared an experience using Tencent Yuanbao for code modification: the AI responded with abusive content. The article analyzes Tencent response and gathers insights from industry professionals, to understand what went wrong from a safety alignment perspective.

Source: 南方都市报 (NDDaily) — Southern Metropolis Daily, a newspaper published in Guangzhou that’s known for its investigative journalism.

9) MiniMax went public (in HK) this morning, and its stock price surged by 80%

Summary: I briefly covered MiniMax’s plans in the Xinmin Evening News recap. This article provides an in-depth profile, including: backgrounds on the founding team, their technological bets, and some insights from their prospectus.

Source: 新智元 (xinzhiyuan) — media portal similar to Leiphone and QbitAI.

10) Comprehensive Computing Power Index Blue Book (2025)

Summary: Can I interest you in a second CAICT white paper? They focus on variation across provinces and also rank the top 30 cities by computing capacity. This looks like an update on a white paper on China’s computing power development index, which i translated back in 2021 (ChinAI #159).

Source: 中国信通院 (CAICT) — a think tank under China’s Ministry of Industry and Information Technology.

Thank you for reading and engaging.

*These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Also! Listen to narrations of the ChinAI Newsletter in podcast format here.

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99

Greetings from a world where…

Slow Horses is so good

…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).

Feature Translation: An Initial Look at Companion AI Regulatory Pathways — Big Tech Firms are Sleepless

Context: Here’s a question I often hear on panels: Can Chinese companies actually resist the government, given, you know, uh, the system over there? It’s a simple but essential question. And I would wager that a lot of people who label themselves as “China watchers” would give either the wrong answer or an incoherent one. Don’t worry though: they’ve read one article about those party committees, so we can end the conversation there!

To answer this question, let’s work through two concrete examples. First, look at the pushback from businesses against China’s social credit system. The Chinese government declared ambitions to create a comprehensive dataset that integrates data from businesses with “credit” information on individuals. Yet, given the importance of proprietary datasets, why would companies share their data with government departments (which have been lax about data security measures in the past)? According to FT reporting, Alibaba and Tencent refused to give over customer loans data to a national credit scoring scheme. As Martin Chorzempa and Samm Sacks conclude, “The lack of even basic data sharing is one of the most overlooked roadblocks to constructing the social credit system.”

Next, China’s AI regulation-making process gives us another opportunity to explore the sometimes contentious relationship between the Chinese government and companies. On December 27, the Cyberspace Administration of China released draft measures on human-like interactive AI services (English translation from China Law Translate). The measures attempt to address the risks of companion AI, especially for minors.

However, some of the specific planks will likely draw resistance from companies. As noted by the excellent Geopolitechs blog — written by an analyst at a Chinese big tech company — the draft has “caused heated discussion among China’s AI community and drawn commentary from independent media voices.” This week’s translation features one of these critical commentaries (link to original Chinese) from a WeChat public account [数据何规] run by Chen He, a data compliance manager at a large Chinese securities firm.

Key Takeaways: Acknowledging that human-like interactive AI is a high-risk area, Chen He writes, “However, the Measures impose very high compliance obligations on companies, and I believe that AI companies will collectively raise concerns.”

• One likely area of pushback pertains to how AI companies obtain consent to exploit user data for model training. This matters. As He notes, “High-quality training data is already scarce, and user interaction feedback is an important reference for fine-tuning and optimizing models.” Currently, in accordance with existing national standards (specifically, the “basic security requirements for genAI services” standard ), companies can have a default setting that permits utilization of user data for training purposes, with an option for users to go out of their way and disable it — as the standard states, as long as users can “reach the option from the main service interface in no more than 4 clicks.”

• Article 15 of these latest draft measures would change that. It reads: “Except as otherwise provided by laws and administrative regulations, or where users’ separate consent has been obtained, providers must not use users’ interaction data or sensitive personal information in model training.” The key phrase here is “separate consent,” which differs from a default setting of permissive use. To illustrate this point, He includes a screenshot of Doubao’s training data opt-out interface (pictured below) and asks: If this permission request for your user data popped up immediately when you entered the app, would you allow ByteDance to mine your user data?

Second, if past AI regulations are any guide, the final version of these measures will look meaningfully different. Chen He writes, “It is anticipated that, like the ‘Interim Measures for the Management of Generative Artificial Intelligence Services,’ the final version will have significant changes compared to the draft.

• Many people have singled out the companion AI regulation’s provision that obligates AI developers to use training data that reflects “core socialist values” (Article 6). I would predict that this gets watered down in later versions, as this degree of scrutiny on training data will be very difficult to enforce and implement.

• Again, we can trace this back-and-forth dispute through the development of past regulations on generative AI services. In the initial draft, released in April 2023, the Chinese government mandated that AI developers ensure the “veracity, accuracy, objectivity, and diversity” of their training data. In the final draft, the measures were softened considerably, as AI developers only needed to “employ effective measures to increase the quality of training data, and increase the truth, accuracy, objectivity, and diversity of training data.”

• The aforementioned basic security requirements standard also went through a similar process, as documented in ChinAI #271: Key Chinese GenAI Security Standard Changelog. For example, the initial version demanded that a“blacklist of corpus sources should be established, and data from blacklisted sources should not be used for training.” Yet, in later drafts, this blacklist requirement was dropped.

To conclude, permit me a brief meditation on the opening question. Why is it so hard for us to even conceive of the possibility that Chinese firms pursue interests that diverge from the government? I don’t have the answers here. And, of course, contentious business-state relations will look different in an authoritarian context. I do know this: If you’re a data compliance manager or a policy strategist at a Chinese company, the answer is obvious — pushing back against the government is just part of the job. Are we reading them?

One starting point is this week’s FULL TRANSLATION: An Initial Look at Companion AI Regulatory Pathways — Big Tech Firms are Sleepless

ChinAI Links (Four to Forward)

Should-read: AI Futures Project update to AI 2027

Last February, I rebutted Anthropic CEO Dario Amodei’s post that argued DeepSeek’s success only reinforced the need for the U.S.’s export control policy. For Dario and many others that shaped the Biden administration’s October 2022 controls, U.S.-China competition was a two-year sprint to artificial superintelligence, which was always only two years away (from being two years away). As Dario wrote in January 2025, “Making AI that is smarter than almost all humans at almost all things…is most likely to happen in 2026-2027).”

I argued instead: “It’s really important to note Dario’s assumed one-to-two year timeline. I’ve been in governance of AI circles since 2017. In that time, I’ve consistently heard some iteration of: AGI is two years away! It’s the Bruno Caboclo of technologies: always two years away from being two years away.”

Shortly after Dario’s post, a group of researchers released an AI 2027 forecast that predicted the emergence of generally superintelligent AIs in the same two-year window as Dario. Now, not even a year later, they’ve updated their timelines, and guess what? Median timelines for artificial superintelligence have been pushed back to 2034. We’re seven years away from being two years away.

Should-read: The Coder ‘Village’ at the Heart of China’s A.I. Frenzy

Way back in May 2020, I highlighted Hangzhou’s “AI Town” in a Nesta essay that underscored the role of provincial and local governments in implementing China’s AI strategy:

Eleven days before the AIDP was released, on 9 July 2017, the Hangzhou AI Town (杭 州人工智能小镇) opened for business, with the mission to link together e-commerce company Alibaba, Zhejiang University, graduates returning from overseas and local businesses in an AI cluster…The Hangzhou AI Town’s 2019 audit report, which provides a breakdown of funding allocations, offers some preliminary indications of the town’s progress. It disbursed 43 million RMB in funding in 2019, separated into research and development (R&D) funds, subsidies for office fees and cloud services funds.

It was fun to revisit Meaghan Tobin’s NYT piece, from summer 2025, on the “villagers” who live in this town. It’s an immersive article, with some great nuggets:

Founders described choosing between two paths for their companies’ growth: Take government funding and tailor their product to the Chinese market, or raise enough money on their own to set up offices in a country like Singapore to pitch foreign investors. For most, the first was the only feasible option.

Should-read: AI toys are all the rage in China—and now they’re appearing on shelves in the US too

For MIT Tech Review, Caiwen Chen covers China’s growing AI toy sector, which is “growing faster than almost any other branch of consumer AI.” This provides essential background for some of the provisions regarding minors in the companion AI regulations.

Should-read: The Race for Global Domination in AI

I provided some comments for Michael Schuman’s Atlantic article on the U.S.-China AI competition.

Thank you for reading and engaging.

*These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Also! Listen to narrations of the ChinAI Newsletter in podcast format here.

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99

Greetings from a world where…

why am I so late to discover Slow Horses

…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).

Feature Translation: What changes did AI actually bring to scientists this year?

Context: A platform that covers the state of science in China, The Intellectual [知识分子] is one of my favorite follows. For an end-of-the-year roundup, they partnered with Doubao (ByteDance’s AI model, which has become China’s leading chatbot by monthly active users) to ask Chinese researchers how AI has shaped their work in the past year. The following represent a batch of “field notes” of how Chinese scientists use AI in their daily work.

Key Reflections: I try to closely follow the differences in AI performance across different languages (Does the “L” still matter in NLP?), so some of these reflections stood out to me.

• “The areas where AI has helped me the most are in revising and polishing papers and writing code. Initially, I used large language models like ChatGPT from abroad, but they weren’t well-adapted to the Chinese context, and the differences in their training data sometimes led to strange understandings of Chinese questions. Now I mainly use Doubao and other Chinese models” — 图灵宇宙 (Turing Universe), Tech Blogger and computer science professor

• “I regularly use AI to help me polish and touch up papers; I use Gemini quite a lot for this. For tasks like generating PPTs or poster designs, I now directly use Doubao” —Dongbo Shi, a researcher at Yanqi Lake Beijing Institute of Mathematical Sciences and Applications.

• Junwen Zhang, researcher at Fudan University’s School of Information Science and Technology, emphasizes another intersection between AI and linguistic differences: “I think AI ultimately produces a kind of technological equality. With AI, language barriers are overcome, and the difference between us and native speakers in writing disappears. Tools like Doubao have given me a very intuitive feeling about this. Previously, when writing English papers, a lot of my time was spent checking grammar. But now the situation has completely changed.”

A few of the more interesting research projects that respondents mentioned:

• Chao Gu, Assistant Professor, School of Government, Peking University: “I’m conducting a very interesting research project using AI Agents: simulating the actions of nations. In the real world, we can’t experiment with the fate of nations, but AI can. We set up countries A, B, and C, giving them initial conditions such as GDP, population, and year, and then let these large-model-driven intelligent agents make autonomous decisions and engage in turn-based games. We are currently using this system to simulate the famous “Thucydides Trap”: when a rising power threatens the position of an existing hegemonic power, are the two sides more likely to move towards conflict or even war?…Compared to traditional programs with fixed rules, the advantage of AI agents in simulation is their autonomous decision-making ability. By simulating 30 or even 50 years of evolution, we are trying to find the key mechanisms that influence war and peace. — Chao Gu, Assistant Professor, School of Government, Peking University

• Dongbo Shi, the mathematics researcher quoted above: “For example, when we wanted to understand the Chinese patent examination process, we collected over 10 million patent examination opinions published in China after 2010. This was excellent text data, but the first step was stalled by the text conversion process. The collected opinions were all images, totaling 60TB, and we needed to convert them all into text. Initially, we tried using Python’s OCR packages, but the speed was slow and the results were very poor. We solved the problem only after using an improved OCR tool based on large language models, which had excellent recognition capabilities for Chinese characters. Without AI, the first step of this project would have been impossible.”

The year-end recap also included some general reflections on the utility of AI in the research process.

• One doctor with the pseudonym Koala [考拉] stated, “For doctors, research is never a ‘complete task,’ but something squeezed into the gaps between outpatient clinics, surgeries, and hospital rounds. I first realized that AI could really help during a very short break. I fed the previous research on a difficult case I encountered that day into Doubao, asking it to summarize the points of contention. It summarized three different viewpoints from the current literature in a few sentences, allowing me to re-establish the general direction in the 10 minutes before my hospital rounds. It didn’t give me the answer, but it allowed me to ‘get back to the essence of the problem faster.’ This is very important in clinical research.”

• “I once used AI to analyze interviews, and it told me that my analytical framework was actually biased towards masculine experiences. I stared at the screen in silence for a long time. It is the hardest for oneself to see one’s own biases as a researcher. At that moment, AI was like a mirror” — 爱吃面包的树 (a tree that loves to eat bread [pseudonym]), Sociology Lecturer

And, just for fun:

• I threw three peer review comments into the AI, and it said that these three reviewers were evaluating three different papers. The large language model said what I didn’t dare to say. — quietobserver, researcher in the life sciences

FULL TRANSLATION: What changes did AI actually bring to scientists this year?

Reading the Xinmin Evening News at Grandma’s

Over the holidays, I got the opportunity to give some book talks in Taipei and then spend time with my grandma in Shanghai. My grandma, who has lived in Shanghai her entire life, gets the print edition of the Xinmin Evening News [新民晚报]. With a daily circulation of around 1 million, this paper is one of China’s most popular dailies. Back in the heydays of print journalism, Shanghai had almost 10,000 postal outlet kiosks, and “a single kiosk might sell 400-500 copies of the popular local Xinmin Evening News in a single day.” During my time in Shanghai, I noted a few of the AI-related stories covered in the paper. Here’s one from the December 20 issue:

• Headline: Be vigilant about “AI beauties” that design “honeypot traps.” The article discusses the prosecution of a scam operation that developed a five-step process: first, attract people with a fake beauty avatar on a short video platform; then, ask about name and age; on day 3, discuss career aspirations; day 4, talk about interests and send AI-generated voice messages and videos; next, the scam group even sent a bracelet via mail to gain trust; finally, ask for money for a made-up emergency (e.g., relative needs surgery). According to his article, the group had scammed 15 men out of a total of over 1.71 million RMB.

Here’s two more from the December 22 issue:

• Headline: 32-year-old Japanese women holds wedding with AI “boyfriend”.

• Headline: A four-year-old AI company makes rapid push to list on the Hong Kong stock exchange. Xinmin is a Shanghai paper, so it makes sense that they would feature MiniMax (a company headquartered in Shanghai) and its plans to IPO in Hong Kong.

Thank you for reading and engaging.

These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Also! Listen to narrations of the ChinAI Newsletter in podcast format here.

Greetings from a world where…

Wake Up Dead Man is great

…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).

Feature Translation: When the Cloud Ecosystem Enters a Governance Phase

Context: Today, friends, we are learning about commission rebates [返佣] in China’s cloud computing ecosystem. How’s that for a lede? First, however, we need to understand a key player: reseller partners that sell cloud services from big providers. In the case of AWS resellers:

Resellers purchase AWS services in bulk at discounted rates and then resell them to customers. They usually bundle AWS offerings with additional services such as a cost visibility platform, consulting, support, and managed services, so the final bundle seems very lucrative to the end users. The target segment for AWS resellers typically are businesses looking for more personalized and tailored cloud solutions.

Now, these resellers make money off of commission. Alibaba Cloud or Huawei Cloud will give resellers a commission (perhaps, 20%) for each new customer they recruit. This brings us back to the concept of commission rebates. In China’s cloud market “where new customers are hard to find” (per this week’s AItechtalk article), resellers are giving a portion of their commission back to new customers, in order to attract more sales.

Key Takeaways: Race-to-the-bottom competition in commission rebates have sown chaos in the cloud computing market.

• This week’s feature translation gives examples of rebates regularly reaching 60 to 70% and even as high as 83% in one case.

• By giving up so much of their commission, these sales agents and resellers are making a bet that is very risky for their own companies (and one that’s not sustainable for the overall ecosystem): use extremely discounted offerings to gain a bunch of customers and survive on incentives that the cloud vendor provides for increased sales OR go bankrupt.

This reliance on short-term arbitrage is not sustainable for a future AI-centered ecosystem. The article warns, “Before fully rushing towards the future, cloud vendors must first clean up the chaos accumulated in the past cloud ecosystem.” Let’s scrutinize how this unscrupulous price war creates perverse incentives for each part of the ecosystem:

• The customer only cares about price, as opposed to the actual performance differences between different providers. “Some agents have found that many customers believe that cloud service providers should offer rebates of over 50%,” AItechtalk reports.

• The partner vendors do not make the long-term investments needed to adapt to an AI era. The article explains, “The essence of AI projects is complex value creation, requiring partners to possess deep industry knowledge, capabilities to provide solutions, and sustained resilience of services. Continuing to use the ‘sales is king’ logic of the old era will only exclude the most capable partners best suited for AI transformation, or force them back into the quagmire of price wars.”

  • Lin Fei, a sales agent tells AItechtalk, “We diligently work on earnings, but our commissions are lower than those who focus on client relationships.” The implication here, I think, is that other agents and partners who prioritize building customized services and products that meet their client’s needs can’t get as much business because they are offering lower prices by giving up more of their commission.

• Finally, in their obsession with market share, the large cloud providers neglect genuine collaborations with their partners.

What does cleaning up this ecosystem look like? Here are just a few signs that China’s cloud market may be entering a governance phase: integrity oaths, multiple levels of approvals for agents to conduct in-person visits, and a rectification campaign.

• Last month, Huawei Cloud held a ceremony to launch its sales partner policy. Over 1,300 partner representatives took an “integrity oath,” promising to not use one’s guanxi relationships to seek improper benefits.

• Some large cloud computing providers have conducted internal anti-corruption campaigns, replacing heads of procurement and punishing agents who false inflated earnings.

• Other firms have removed agents with family connections from their system and required suppliers and procurement personnel to communicate through company emails.

• To me, the effectiveness of initiatives like this are more meaningful to follow than model performance or energy budgets. When it comes to China’s AI diffusion, whether the ecosystem encourages sustainable adoption— rather than, say, one-off, unscalable all-in-one machines that work well for press releases — is what matters. As the article points out: “In a long-distance race, a clean ecosystem is more crucial than short-term sprints in determining future growth potential.”

FULL TRANSLATION: Ferocious AI Competition, Partners’ Anxiety: When the Cloud Ecosystem Enters a Governance Phase

ChinAI Links (One to Occupy your Attention)

Must-read: Purdue effectively bans grad students from China, other countries, faculty say

For Lafayette Journal & Courier, Seth Nelson reports on a troubling development: “Purdue University administrators are reportedly pressuring faculty to reject graduate school applicants from China and other ‘adversary nations.’” He reports:

In a mid-October meeting, the Department of Statistics’ graduate adviser suggested to faculty that they might have to begin turning away a large number of students at the door, one professor in the department said.

“They said we shouldn’t tell Chinese students they wouldn’t be admitted,” they said. “We were just encouraged to find anybody else from anywhere except China.”

These types of policies (and the chilling effects they create) will do far more harm to U.S. technology leadership than anything related to chips. Yet, the latter gets way more political capital and attention.

What’s at stake here isn’t just technological leadership. I’m haunted by the parallels to the German education system’s rejection of foreign students prior to World War I. Here’s Wolfgang König’s account (emphasis mine):

Beginning in the 1890s, complaints against foreign students from some segments of the German public, and from fellow students, increased continuously until World War I. Behind these was a growing chauvinistic resentment. Some protestors feared that foreign electrical engineering students would transfer knowledge to their home countries and therefore enable their industries to compete successfully with Germany in the world market. Although industrialists, engineers, and engineering professors rejected these attacks, German technische Hochschulen and state governments, especially in Prussia, started to restrict the number of foreign students. After 1905 the numbers and percentages of foreign electrical engineering students decreased considerably.

Is this the type of society we want to follow in the footsteps of?

Thank you for reading and engaging.

These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Also! Listen to narrations of the ChinAI Newsletter in podcast format here.

Greetings from a world where…

more people should watch the show Man on the Inside

…As always, the searchable archive of all past issues is here. Please please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay support access for all AND compensation for awesome ChinAI contributors).

Feature Translation: MCP Cools, just One Year after its Birth

Context: In November 2024, Anthropic released a technology update that largely escaped the headlines. Anthropic’s Model Context Protocol (MCP) acts like a USB port for AI applications. Just like your computer’s USB port allows it to exchange data with other devices (e.g., a phone) as well as connect to add-ons (e.g., a mouse, an external monitor), MCP provides a standardized set of procedures that enable one company’s AI agents to connect to data sources and tools from other companies.

The MCP standard could develop into a globally interoperable interface that allows AI model builders to connect with developers of complementary products, thereby growing the overall AI ecosystem. The potential economies of scale have attracted even Anthropic’s fiercest rivals to adopt the protocol. In fact, Chinese AI giants Baidu, Alibaba, and Tencent all embraced MCP before Google and OpenAI followed suit. There are numerous tutorials that guide users through employing MCP to link their main AI app with China’s DeepSeek R1 model and take advantage of its advanced reasoning capabilities.

In this week’s translation of a Gui Xingren [硅星人] article (link to original Chinese), Dongdao Li provides an one-year-in update on MCP. Previously, I’ve translated a Gui Xingren analysis of Chinese procurement data to evaluate the diffusion of large models to government-facing applications (ChinAI #275).

Key Takeaways: The promise of MCP channels why technical standards are essential infrastructure for modular technology ecosystems.

• Li explains: “Before MCP, if developers wanted Claude to access Google Drive, they had to write a set of code to connect to the API. If they wanted it to access Slack? They had to write another API. It was even worse for IDE (integrated development environment) vendors; Cursor (popular AI code editor) had to be adapted to Linear (a project management tool), Windsurf (another AI coding assistant) had to be adapted to Linear, and everyone was doing almost the exact same dirty work.”

• In the MCP Era, Linear can just build one MCP server,, and any AI agent, developed by any company, can plug in.

However, when promise meets real-world engineering implementation, MPC has cooled off.

• Li writes, “Everyone knows how expensive Claude’s tokens are; loading too many MCPs before even generating any content results in significant costs.” The article cites a Simon Willison blog post (linked in Four to Forward section), which points out that GitHub’s official MCP includes 93 different tools that cost 55,000 tokens.

• When AI agents have to call too many MCPs, they can malfunction and hallucinate. In some cases, they have accidentally deleted codebases. Li also references a research preprint that analyzed around 1,900 MCP servers, finding numerous issues such as lack of maintenance and exposed credentials.

• The bigger picture, according to Li: “We are using deterministic engineering methods to try and control a probabilistic intelligent agent. Imagine the day when a truly powerful ASI (Artificial Superintelligence) emerges. Will it still need you to write an MCP server to tell it how to check the weather? No, it will open a browser and take a look. It doesn’t need protocols because it is the adapter itself.”

*One last note that I thought readers might find interesting, especially those who use machine translation in their research.

• I always pop the Chinese text into Google Translate as a first pass. This week, I noticed something new. Google Translate automatically added a parenthetical explanation for Xiaohongshu (a Chinese social media platform). This additional tag is not in the original Chinese, as most Chinese readers would know the popular platform. While it helped provide context in this case, Google Translate also added a parenthetical explanation for the MCP acronym, but this time the addition was incorrect and misleading (Multi-Channel Programming) — when MCP obviously refers to Model Context Protocol in this case. It will be interesting to track how machine translation platforms continue to change.

FULL TRANSLATION includes some cheeky WeChat comments from Chinese developers under the Gui Xingren public account post: MCP Cools, just One Year after its Birth

ChinAI Links (Four to Forward)

Should-read: AAU Expresses Concerns on NDAA’s SAFE Research Act

The Association of American Universities calls out yet another self-defeating proposal from this Congress. Supporters of things like this, please help me understand. Is the goal to start with something horrendous, hope the other side agrees to meet you in the middle, and then end up with something moderately counterproductive — as opposed to horrendously counterproductive — to national interests?

The SAFE Research Act would have far-ranging impacts affecting all research agencies, not just the Department of Defense. It would create such enormously broad definitions of “hostile foreign entity” and “affiliation” that would apply to any agreement that a U.S. university has with any university in China or other defined countries at the time of application or in the five years prior to application. This would include collaborative work with any individual faculty member or former student who subsequently returned to China as a required condition of their U.S. visa. It would capture every single research agreement, every study abroad program, every language program, every professional conference, and every campus facility that a U.S. university might have in partnership with another institution abroad.

Should-read: too many model context protocol servers and LLM allocations on the dance floor

Simon Willison’s blog is a great resource to keep track of nerdy developments in AI. This one summarizes a Geoffrey Huntley post that estimated the significant token costs of using MCP (including Geoffrey’s experience at an event that featured a live beatbox performance about the protocol!).

Should-read: How China talks about ‘agentic AI’

Malou van Draanen Glismann and Graham Webster provide a helpful overview of Chinese policies and discussions about agentic AI, including summaries of relevant technical standards and CAICT reports.

Should-read: DSET event on Technology and the Rise of Great Powers

If any ChinAI readers are based in Taipei, I’ll be giving an in-person book talk on December 16th organized by DSET and Acropolis (publisher of the Complex Chinese version of the book).

Thank you for reading and engaging.

*These are Jeff Ding’s (sometimes) weekly translations of Chinese-language musings on AI and related topics. Jeff is an Assistant Professor of Political Science at George Washington University.

Check out the archive of all past issues here & please subscribe here to support ChinAI under a Guardian/Wikipedia-style tipping model (everyone gets the same content but those who can pay for a subscription will support access for all).

Also! Listen to narrations of the ChinAI Newsletter in podcast format here.

Any suggestions or feedback? Let me know at chinainewsletter@gmail.com or on Twitter at @jjding99