The 'Digital China' Plan, cross-border data, ChinaGPT

and

Mar 06, 2023

Welcome to this edition of the DigiChina Update, written by Editor-in-Chief Graham Webster (GW) and Student Editor Tianyu Fang (TF).

If you have not already, please consider subscribing to this newsletter:

You can also keep track of DigiChina following us on Twitter at @digichn. And if you like what you see, please consider sharing this issue with friends and colleagues:

A new ‘Digital China’ Plan, translated

Ever since people started writing in Chinese about our DigiChina Project, it has been a bit awkward that the usual Chinese-language translation is 《数字中国》(“Digital China”), which also happened to be one of the Chinese government’s tech policy buzzwords beginning around the time of the establishment of the Leading Small Group for Cybersecurity and Informatization in 2014.

Digital China, the concept that shares our modest effort’s name, has in recent years moved to the center of official Chinese discourse on technology. Premier Li Keqiang gave it special emphasis around the 2021 unveiling of the 14th Five-Year Plan, in which it represented one of the 19 overall parts. Now, the Communist Party and the State Council have issued a “Plan for the Overall Layout of Building a Digital China.” The full text of the Plan is not apparently public, but DigiChina has translated the fullest state media report.

DigiChina readers will perhaps unsurprisingly be familiar with much of the Digital China Plan content. Specialists will find a close read rewarding, but two things stand out.

First, this Plan has been issued by the Communist Party of China, as well as the State Council. This stands in contrast to the previous major tech sector plan, the 2017 New Generation Artificial Intelligence Development Plan (AIDP), which was a State Council product, mainly of the Ministry of Science and Technology system. The Party’s role is emphasized in the new Plan:

“Uphold and strengthen the overall leadership of the Party in the building of a Digital China; under the unified and centralized leadership of the Party Central Committee, the Central Commission for Cybersecurity and Informatization [CCCI] is to strengthen overall coordination, comprehensive promotion, and supervision and implementation of building a Digital China.”

Given that the Cyberspace Administration of China is the office of the CCCI and is a one-institution, two-nameplate Party-state hybrid organization—and in light of a WSJ report that China may create a new entity for data governance—we’ll be watching for signs and implications of any shift in Party vs. state roles here. As for other outcomes to track, note that unlike the AIDP, which was released in full and included some fairly concrete metrics against which progress could be measured, much of the Digital China Plan as released is too vague to specifically assess.

Second, in the grand tradition of Chinese planning, we have a new numerical framework: 2522. Two foundations, five integrations, two capacities, and two environments. Study up!

The Plan clarifies that the construction of Digital China will be laid out in accordance with the overall framework of "2522," that is: to use the "two foundations" of digital infrastructure and data resource systems; to promote the deep fusion of the “five integrations” of digital technology with the economy, politics, culture, society, and ecological civilization; to strengthen the "two capacities" of the digital technology innovation system and digital security shield; and to optimize the "two environments" of domestic and international digital development.

(For the record, when naming DigiChina circa 2017, we did not consider that it would likely translate as 数字中国, a term that was already circulating, though much more quietly than “cybersecurity and informatization.” It seems we’ll increasingly have some explaining to do.) –GW

📄 DigiChina translation: Plan for the Overall Layout of Building a Digital China
🀄️ CAC Deputy Director Cao Shumin writes in Red Banner Manuscripts: “Driving Chinese-style Modernization through Informatization and Digitization” 以信息化数字化驱动引领中国式现代化
📋 Pacific Forum has a well-timed paper diving into the “Digital China” concept and arguing it is the world’s first digital grand strategy.

China’s cross-border data system grinds into motion

Since the Cybersecurity Law was finalized in 2016, businesses with any cross-border activity have been waiting to see precisely how the law’s provision governing outbound transfer of certain data (Art. 37) would take shape. Long-awaited measures fleshing out the implementation of this and subsequent cross-border data provisions finally took effect last September, with enforcement for pre-existing data flows taking effect March 1 this year.

Multinational companies reportedly rushed to figure out whether their operations in China were covered and, if so, to complete the filing and review process in time. In January, a cancer research collaboration with a Dutch medical center was the first to have its data transfers approved, Caixin reported, with Air China getting the second approval.

Relatedly, the China Cybersecurity Review Technology and Certification Center was the first institution approved to conduct personal information protection certifications under the Personal Information Protection Law, WilmerHale wrote. And CAC published Measures for Standard Contracts for the Outbound Transfer of Personal Data, as well as a template contract, another long-awaited piece in the cross-border data protection puzzle. –GW

📄 China Law Translate: Translation of the Measures for Standard Contracts for the Outbound Transfer of Personal Data

OpenAI’s ChatGPT is not so open in China

In China, the popularity of ChatGPT has led to substantial public interest in the progress of its own generative artificial intelligence. But if you’re physically in China, it’s a hassle to even poke around with ChatGPT itself: the San Francisco startup OpenAI has banned Chinese (+86) and Hong Kong (+852) numbers for account registrations—meaning although anyone in the PRC can access the OpenAI website that hosts ChatGPT, they can’t use the tool unless they can receive an SMS verification code sent to a non-Chinese phone number. And China isn’t the only country where OpenAI services are restricted: Russia, Vietnam, and Saudi Arabia are not among the 162 supported countries and territories.

Tech-savvy users in China have come up with workarounds, using foreign virtual private networks (VPNs) to bypass IP checks and U.S. VoIP services for SMS verification. They also need to make sure that the frequent switching of VPNs will not trigger OpenAI’s anti-abuse firewall. Some developers have built bots on WeChat and other China-based social media networks using OpenAI’s API, but a handful of them have been suspended by Chinese authorities and their host platforms.

By character count, only some 0.19% of GPT-3’s training dataset is in Chinese, compared to 92% in English. In this context, it’s surprising how well ChatGPT works in Chinese, though still worse than its English-language capabilities. Still, the question remains: Why hasn’t China invented its own ChatGPT? It’s convenient to attribute the Chinese lag in generative AI to its political system and top leader Xi Jinping’s critical view of consumer technology, but that doesn’t explain China’s tremendous progress in AI in recent years. As Louise Matsakis reports for Semafor, regulatory risks, content moderation concerns, U.S. export controls, brain drain, and the priority of profit-driven research are likely contributors to the relative stagnation.

It is, however, very likely China isn't that far behind. Baidu, which has released an image generator and is planning to launch its own chatbot Ernie, might integrate its own version of GPT into Baidu AI Cloud, as Kevin Xu argues in

Interconnected

. Other companies including Alibaba, Bytedance, JD, and Qihoo 360 are entering a generative AI arms race alongside Baidu. –TF

What we’re reading

Lessons from EU and Chinese approaches to AI governance. Matt O’Shaughnessy and Matt Sheehan compare AI regulatory regimes—the vertical approach of the AI Act in Europe, which places constraints on each category, and the horizontal approach of China’s algorithm registry.

🔍 Carnegie Endowment: Lessons From the World’s Two Experiments in AI Governance

More on EV primacy in China. Chinese authorities are piloting a program that deploys new energy vehicles to public-sector services, including taxis and buses as well as sanitation, postal, and logistics vehicles.

An Internet of Vehicles. Meanwhile in Jiangsu province, the city of Wuxi has issued guidelines for citywide adoption of “internet of vehicles” technologies—building infrastructure connecting networked cars—that go into effect on March 1. Last February, the MIIT issued guidelines for cybersecurity and data security for the “internet of vehicles,” which, we can be honest, is a better coinage in Chinese.

📰 Sixth Tone: In a First, Chinese City Unveils Law for Internet of Vehicles
📄 National People’s Congress: 《无锡市车联网发展促进条例》3月1日施行
📄 Wuxi People’s Government: 无锡市车联网发展促进条例
🔍 Grid: How China built the world’s biggest EV charging network - and left the US far behind

Back in the game. The National Press and Publication Administration (NPPA) approved 87 video game licenses in February—some of them were granted to major players Tencent, NetEase, and Bytedance. It’s among several signs that the video game industry is, at least tentatively, out of the doghouse.

–TF