Can Europe and China Cooperate on AI?

Recently, an Oxford University seminar attracted participation from universities, think tanks, and AI companies from China, the UK, and the EU. The theme was "Can Europe and China Cooperate on AI?"—a topic significantly relevant to China's current AI industry.

Both Chinese tech giants and startups are seriously considering expanding internationally. According to the “February 2024 Global AIGC Industry Monthly Report” from Feifan Industrial Research(非凡产研), the number of China’s outbound AI products continues to grow, with three Chinese applications making it into the global Top 50 AI products list.

China's tech companies traditionally viewed the U.S. market as most promising. However, due to geopolitical risks and the competition between China and the U.S. in AI, the U.S. market doesn’t seem as attractive as before. Despite this, the larger pool of paying users, looser regulatory environment, and more talent in the U.S. still draw many Chinese AI startups, such as 01.AI, Moonshot AI, and MiniMax. These companies currently focus on virtual chatbot services that require minimal computational resources and data, and they've made modest gains. For example, according to SensorTower, MiniMax's Talkie is now the 12th most downloaded AI app globally, following closely behind its U.S. competitor, Character.ai.

But how about Europe as another important overseas market? This is a question I’ve been very curious about, which led me to attend the Oxford event and exchange ideas with some industry professionals.

In Europe, the UK appears to be a prime destination for Chinese AI expansion. Despite the National Day holiday, several Chinese companies travelled long distances to attend the conference, showing their interest. After speaking with a few experts, I found that several factors drive the UK's appeal:

First, since Brexit, the UK has maintained a certain regulatory distance from the EU regarding cross-border data and AI regulation, creating a more favorable environment.

Second, English, as the primary language for pre-trained data, gives the UK a natural advantage over countries like France and Germany, which also aim to attract AI investment.

Lastly, some companies are optimistic about the current Labour government's China policies, seeing them as less extreme and thus offering short-term manageable geopolitical risks.

Due to compliance costs under GDPR and the AI Act, Chinese companies, particularly startups, remain hesitant about entering the EU market. Experts from Chinese AI firms universally questioned the scientific basis of the EU AI Act, arguing that the so-called “Brussels Effect” essentially uses vague laws and principles to hinder technological progress. Some experts pointed out that AI develops rapidly, with risks gradually emerging. Therefore, regulatory frameworks should be based on empirical research rather than excessive fines and outdated laws that stifle innovation.

Beyond regulation, expanding into Europe involves other challenges like AI talent availability, visa policies for tech workers from China, computing power, and the costs associated with building data centers.

Although most Chinese AIGC companies are still evaluating the European market, companies like Huawei and Alibaba Cloud have long established their AI infrastructure in Europe. Huawei has been promoting cloud-based AI platforms in Europe in recent years, aiming to expand its AI ecosystem's influence. Alibaba Cloud already has five data centers in Europe and has built local teams in Germany, the UK, and France, offering AI and machine learning solutions for various industries like automotive, retail, and gaming.

For well-known reasons, European AIGC companies have not entered the Chinese market, but some have attempted to collaborate in AI hardware and AI-assisted drug development. For instance, in 2021, British AI chip maker Graphcore partnered with Chinese tech giant Digital China. Digital China became Graphcore’s distributor in China, helping build an AI distribution and channel network. Another example is Sanofi, a French pharmaceutical company, which signed a $1.2 billion deal with Chinese biotech firm Insilico Medicine to use its Pharma.AI platform for drug discovery.

Compared to the uncertainty surrounding Chinese AI companies expanding into Europe, I was quite surprised by the collaboration in AI fundamental research between China and Europe.

An expert from the Mercator Institute for China Studies presented findings from an empirical study showing that between 2017 and 2022, joint AI research between Europe and China steadily increased. Chinese and European AI researchers co-authored 36,951 English papers, second only to joint research between China and the U.S. In 2021, these collaborations accounted for one-tenth of all AI papers produced in Europe.

China primarily learned from and imported advanced European technology in past high-tech collaborations. However, in the AI field, China now leads the EU overall. According to a study by Georgetown University's Center for Security and Emerging Technology, China surpassed the EU in AI research paper citation rates as early as 2016, becoming one of the top countries in the world. China significantly outpaces the EU in computer vision research and is gradually surpassing it in natural language processing and robotics.

Take, for example, the co-authored AI papers between China and Europe. Many European authors were Chinese nationals studying or conducting postdoctoral research in Europe. Among the papers that disclosed funding sources (about half of the total sample), 80% were funded by China (60% from Chinese government projects and the rest mostly from Chinese universities). Almost half of the China-Europe AI research collaborations involved Chinese and British researchers, with Chinese researchers contributing to 21% of the UK's AI papers in 2021. In contrast, British researchers contributed to only 3% of Chinese AI papers during the same period. This indicates that China is making a larger contribution to the UK's AI research, with the UK benefiting more from China.

According to the said MERICS research, China-EU AI investment has seen growth since 2017, peaking in 2021.

Overall, Chinese investment in Europe’s AI industry remains limited, but European investors have shown a higher level of engagement in China’s AI sector. According to data from the French Institute of International Relations, between 2019 and 2023, the majority of Europe’s tech investments in China were concentrated in AI, with Germany and France being the most significant contributors. By the end of last year, Europe’s investments in China’s AI industry accounted for about 4% of China’s total AI funding, with companies like Siemens and Volkswagen establishing AI labs in China.

Between 2017 and 2022, 64 Chinese AI companies raised around €11.16 billion from European investors. For example, the UK-based Aurora Private Equity was a key investor in Chinese tech company Marknum Technology’s €12.8 million Series B funding round. Similarly, since 2019, Bosch Group's venture capital arm Robert Bosch Venture Capital has been investing in China’s AI sector, with notable investments in companies like WeRide, which focuses on L4 autonomous driving, and Xingge Technology (a subsidiary of Cambricon), which specializes in autonomous driving chip development.

On the other hand, Chinese investors have made 1.2% of total AI investments in Europe, amounting to about €1.4 billion. The two largest investments went to Exscientia, a UK-based AI-driven drug discovery company, and Agile Robots, a Germany-based company developing AI-powered industrial robots. Additionally, French startup Prophesee, which specializes in neuromorphic vision systems, received investments from Chinese firms like Sinovation Ventures, Xiaomi, and Inno Chip (a Shanghai-based semiconductor venture capital fund).

Overall, the momentum for China-EU collaboration in the AI field remains positive, but many experts at the Oxford seminar pointed out that the geopolitical competition between China and the U.S. has significantly impacted AI cooperation between China and Europe.

Taking the UK as an example, despite having world-leading AI research institutions like Oxford, Cambridge, and the Turing Institute, the country lags significantly in the application and commercialization of technology. This has led many UK universities to be open to collaborating with Chinese AI companies. However, university experts indicated that since the UK government views AI as sensitive technology, there is extra caution when working with Chinese companies, with risk assessments taking a long time. Some UK universities that rely heavily on funding from defense and security sectors have been forced to reduce or even cancel collaborations with Chinese universities.

It was mentioned that different UK government departments have varying attitudes toward AI collaboration with China. The trade people are relatively encouraging it, while the defence, security, and foreign service people are highly cautious and inclined to highlight the risks. Some UK venture capital firms and universities have already received alerts about the risks of collaborating with China.

Under the UK’s 2022 National Security and Investment Act, foreign investment scrutiny, especially regarding Chinese investments in high-tech sectors like AI, has become increasingly stringent. In 2022, the UK government blocked Manchester University from licensing a visual sensor technology to Beijing Infinite Vision Technology Company, and it also reviewed and rejected a deal for Hong Kong-based Super Orange HK Holding Limited to acquire UK chip design software company Pulsic Limited.

The EU is advancing its own economic security strategy and emphasizing "de-risking" from China, thus focusing more on "research security." It was estimated that the EU will likely conduct stricter reviews of end-users of China-EU joint AI research in the future, which could limit AI research cooperation. On the business side, some European companies, fearing reputational and political consequences, may voluntarily sever ties with Chinese AI investors and cloud service providers.

Additionally, the EU's emphasis on digital and technological sovereignty cannot be overlooked. Given China’s technological advantages and the scale of its AI ecosystem, the EU is concerned that AI cooperation might eventually become one-sided, with China benefiting more, draining Europe of its AI talent, or leading to Chinese AI companies, much like U.S. firms, becoming too dominant in Europe.

Nevertheless, the prevailing view among European experts seems to be that Europe’s economic security strategy and de-risking efforts differ from the U.S.’s decoupling from China. Europe is not seeking to curb China’s technological progress but rather aims to manage risks responsibly through cooperation, minimizing risks in key areas while maintaining broader collaboration with China.

However, some have pointed out that with the intensifying AI competition between China and the U.S., the world might eventually see two distinct AI ecosystems, narrowing the space for European AI research institutions and businesses to collaborate with both China and the U.S. simultaneously.

In discussions, European experts expressed disappointment at the lack of dialogue between China and the EU on AI governance, seeing it as an area where China-EU AI cooperation could be strengthened.

Based on their statements, it is clear that traditional concerns such as privacy, data security, and consumer protection remain critical issues for European governments and civil society when it comes to AI technology and its applications. Their focus on AI governance is serious, often even surpassing concerns about the industry's growth itself.

Some experts believe that China, as the world’s second-largest AI power after the United States, has developed its own distinct practices and positions on AI governance, which are closely watched by others. Over the years, China has indeed made significant strides in AI governance. However, there has been relatively little external communication about these efforts.

For instance, China has proposed a global AI governance initiative, but many European countries find it vague, with some terms open to interpretation. Europe, with its strong focus on AI governance, is eager to engage with China in this area. Yet, it seems that China is more interested in dialogue with the U.S. on this topic, paying less attention to Europe. Strengthening dialogue on AI governance between China and Europe would help depoliticize their cooperation in AI and build mutual trust in this field.

During the meeting, a Chinese academic expert responded to concerns about China’s global AI governance initiative, explaining that the initiative is primarily a set of principles, similar to the “Belt and Road Initiative”. It is more of a framework that can encompass many things, and as more countries participate and further practical experiences are gained, the initiative’s content will become richer and clearer.

I share the European experts' concerns. In recent years, China has proposed various initiatives in areas like data security and AI governance, showing its leadership in global governance of emerging technologies. These efforts are commendable. However, China should not stop at issuing documents and reiterating "China's position." Instead, it must listen to, understand, and incorporate the views and proposals of other countries. Otherwise, it may remain only a "Chinese position" and fail to become "global consensus."as the Chinese government wishes.