Disconnecting the Gordian Node

The technology sector is fragmenting. Once a driver of mutually beneficial globalization, technological progress has become a race for geopolitical dominance. States worldwide are “de-risking” supply chains of essential technologies such as semiconductors, upending intricate global production processes. Autocratic regimes are using digital tools to oppress citizens at home and destabilize democracy abroad. And at the frontier of digital technology, China and the US are competing for dominance in artificial intelligence (AI), while experts warn of the global risks of unregulated AI development. As states increasingly use technology to gain dominance over their geopolitical rivals, these new trends of tech weaponization and disintegration have repercussions for international security.

The current chokepoints in semiconductor supply chains are prone to weaponization. States are using these chokepoints to control their rivals’ access to advanced chips. For example, the US and its allies use export controls to deny China the advanced semiconductors it needs for supercomputing and AI technologies to improve its military capabilities.^[1] Since 2019, Japan, the Netherlands, and the US have imposed increasingly strict controls on the export of cutting-edge chips and the materials needed to make them, such as high-end lithography machines, to China.^[2] At the same time, Western countries are increasingly engaging in de-risking – that is, diversifying supply chains to decrease dangerous dependencies. Whether rivals could ever untangle their mutual dependencies, however, remains unclear. Through the Chips Act and the CHIPS and Science Act, the EU and US are respectively investing heavily in domestic semiconductor manufacturing.^[3] These efforts have been nominally successful, as chip manufacturers have announced openings of new fabrication facilities, or fabs, in Europe and the US.^[4] Nevertheless, fab capacities remain highly concentrated in Taiwan, in the shadow of China’s increased military posture.^[5]

Experts have criticized these efforts for ignoring chokepoints beyond the fab dependency on Taiwan. China controls significant shares of global gallium and germanium production, minerals required for chip production (Figure 8.1). It also holds 28 percent of global facilities for assembly, testing, and packaging (ATP) of chips.^[7] Hence, new fabs alone will not secure chip supplies. Their costs, however, are significant. The EU and US chips acts have respectively injected 43 billion and 53 billion US dollars into the chip industry.^[8] China and South Korea have invested even more; thus, the risk of a subsidy race looms large.^[9]

The promise of the internet as a force for global democracy has waned.^[16] Instead, autocracies have weaponized digital infrastructure, using it to control their own populations domestically and create and exploit dependencies internationally.^[17] China has expanded its influence on global digital infrastructure through investments in 5G infrastructure and subsea data cables as well as proposals for new technical standards that would facilitate internet surveillance.^[18] With the Beijing-based company Bytedance, which owns TikTok, it now has a globally succesful tech giant, to whose user data the Chinese government may have far-reaching access through its sweeping Data Security Law.^[19] Meanwhile, Russia is conducting online disinformation campaigns to destabilize democracies, exploiting the propensity of social media platforms to polarize.^[20]

Leading democracies had long ignored these threats. The EU and US welcomed affordable Chinese digital hardware to improve their physical digital infrastructure and relied on lax regulations regarding data security and harmful digital content.^[21] This attitude finally changed in the past decade. Since 2018, the US and other democracies have restricted the use of Chinese hardware in their 5G networks over espionage concerns, though with varying determination.^[22] The EU has established itself as a pioneer in regulating tech platforms to protect its citizens’ data and control harmful digital content.^[23] The US has been reluctant to follow, but concerns about TikTok may signal a mood change.^[24] Finally, democracies have tried to formulate a democratic alternative to China’s autocratic digital vision, for example in the 2022 Declaration for the Future of the Internet.^[25]

However, the diverging approaches of the EU and US to tech governance complicate efforts to realize such an alternative vision. This was discussed at the 2023 MSC Roundtable on Cybersecurity in Brussels.^[27] There, European and US participants alike emphasized the need for transatlantic tech cooperation, but admitted that challenges remain: while the EU worries that unregulated digital technologies will expose European citizens to data security violations and harmful content, the US fears ramifications of the possible regulations on its tech sector.^[28] Moreover, experts from both sides of the Atlantic have pointed out the need to “security-proof” new tech regulations.^[29] The EU and US can ill afford disagreements as they aim to present a joint democratic alternative to China’s digital authoritarianism. The establishment of the US-EU Trade and Technology Council and the 2021 OECD digital services tax agreement demonstrate, however, that they can find mutually beneficial solutions.^[30]

Meanwhile, many states in the so-called Global South refuse to take a side in the digital autocracy–democracy rivalry, as they prioritize development goals over value-based considerations.^[31] They might align their tech regulations with the EU’s to ease trade, while investing in affordable Chinese-built digital infrastructure.^[32] Hence, if the EU and US want to win support for their vision of global digital governance, it must include the development of digital infrastructure in the Global South.

AI technology will be a key determinant of geopolitical power in the coming decades. The abilities of modern machine-learning systems, a subtype of AI, to analyze masses of data at superhuman speed could revolutionize military capabilities.^[33] In 2020, AI beat a US Air Force fighter pilot in simulated air combat.^[34] War games have shown AI-powered autonomous drone swarms to be essential for successfully defending Taiwan against a potential invasion.^[35]

China and the US are vying for dominance in AI, but who is ahead is unclear. National-level metrics paint an ambiguous picture; in 2021, 40 percent of AI journal publications stemmed from Chinese institutions, while only 10 percent came from US institutions.^[36] The US, however, leads China in annual funding of new AI companies, with 542 new AI start-ups funded in 2022 versus China’s 160.^[37] Even Europe outperforms China on this metric, with 293 start-ups in the EU and UK combined. Indeed, despite Chinese and US dominance, the AI race is not just a bilateral one; the EU, the UK, India, and Israel also perform well in several metrics.^[38] Finally, the controls imposed on semiconductor exports to China will likely impede China’s AI progress.^[39]

An unregulated AI race carries risks of its own. AI systems have already caused various incidents, and this problem is worsening (Figure 8.2). Realistic image- and video-generating models are aggravating the disinformation problem.^[41] AI-powered job-recruitment and fraud-detection systems have displayed racist or sexist biases.^[42] Malicious actors could abuse generative AI systems to acquire instructions for making chemical weapons, for example.^[43] Finally, potential military applications of AI also carry security risks. AI models sometimes act unpredictably, which can have fatal consequences if applied to a weapon system.^[44] AI weapon systems with limited or no human oversight also raise questions regarding accountability for the potential war crimes that such systems could commit.^[45]

While policy-makers focus on winning the AI race, they risk underestimating these issues.^[46] Fortunately, several states and institutions are now taking initial steps to regulate AI. Continuing the EU’s digital regulation efforts, the European Parliament and EU member states agreed on the new AI Act in December 2023.^[47] The act classifies AI applications into risk categories and subjects them to restrictions accordingly. In October 2023, US President Joseph Biden issued an executive order aimed at regulating AI. Its provisions announce standards for testing new AI systems, the results of which developers must share with the federal government.^[48] And an international AI Safety Summit held in the UK in November 2023 resulted in the Bletchley Park Declaration, in which states committed to cooperate to address AI risks.^[49] Notably, this declaration was underwritten by China, the EU, and the US. These guidelines stil need to lead to concrete international regulations.

Throughout the tech sector, global cooperation has given way to geopolitical competition. In semiconductor and AI policy, Chinese and US policy-makers are focusing on outperforming each other, and the potential for mutual absolute gains is shrinking. Meanwhile, a confrontation between democratic and autocratic visions on digital governance is unfolding, where China, the EU, and the US are using digital regulation and infrastructure to export their conflicting visions. As omnipresent as this logic of geopolitical competition is, its limits are apparent. Despite de-risking, the complete disentanglement of mutual dependencies between geopolitical rivals in the semiconductor supply chain is hardly possible. In many states in the so-called Global South, development logic still prevails over geopolitical considerations. And finally, even in the geopolitical race for AI leadership, the moral imperative to establish global regulations to mitigate the risks of AI is evident. States thus need to find an equilibrium between unavoidable competition and indispensable cooperation.