While Europe is forging ahead with strict AI rules, China is pursuing a pragmatic course – with strong ideological control. This model could be particularly attractive for governments in the Global South, writes Julia Fiedler. Because even though the EU has ten times as many well-intentioned regulations, China is far more determined in one area: censorship.
Donald Trump is toying with the idea of withdrawing China’s PNTR status. PNTR stands for Permanent Normal Trade Relations and guarantees most countries in the world easy access to the US market. Withdrawing the status would put China on a par with North Korea and Russia, explains Jörn Petring. The economic consequences would also be enormous for other countries.
AI export quotas divide the world into the good guys, the bad guys, and those who have to choose sides, write Antonia Hmaidi and Wendy Chang in today’s Opinion. They argue that the United States is trying to regulate technology while it still can.
Differing values, geopolitical competition, and the race for tech dominance: There are many obstacles to global AI governance that creates a framework for the ethical, security, and legal challenges posed by the technology. At the AI Action Summit in Paris earlier this week, 60 countries signed a final declaration calling for the initiation of a global dialog on AI governance, including Germany and China. The USA did not sign the declaration.
Who shapes the global regulations is crucial. China has already offered to take a leading role. In order to deal with AI together in the competition for influence, it can be helpful to understand how China’s AI legislation works, says Dr. Daniel Sprick, a research associate at the Institute for Chinese Legal Culture at the University of Cologne.
Standards play a central role here. These often primarily regulate aspects of connectivity or interoperability of technical products. However, China also uses standards as a form of sub-legal regulation to ensure that companies act in accordance with the rules. It is already able to export parts of its governance system that are shaped by its own authoritarian ideas.
Innovation needs freedom, but when it comes to artificial intelligence, there is great concern about misuse and harmful effects on society. So how do you draft a law for artificial intelligence that restricts AI without slowing down the breathless technological sprint? In the EU, the AI Act came into force in August 2024, the world’s first comprehensive law to regulate AI. While praised by some, it has also been criticized for its detailed and strict regulations. These could hinder innovation.
While the European regulation comprises almost 500 pages, the current Chinese legislation on AI does not even have 50 paragraphs in total. A comprehensive AI law 中华人民共和国人工智能法 is only being discussed in academic circles. This is despite the fact that regulation in China should actually be particularly strict – after all, even a large language model has to be true to the line. It is not allowed to provide answers on politically sensitive topics.
The fact that there is still no comprehensive AI law in China and that the few legal regulations are so meager may come as a surprise, but it is typical, says Sprick. Socialist law usually pours the program into a law first and then makes adjustments. Step by step, it is examined which further rules are useful – and these are then supplied in the form of standards.
The foundation stone was laid by the Social Credit System with its assessment standards. There are now numerous standards, some of which are also relevant for international consumers. For example, for cross-border e-commerce or data security.
In the field of AI, the Chinese government is clearly still unsure of where it wants to go. This can be seen from the fact that regulations are currently only laid down at the level of administrative guidelines and often have titles that begin with “provisional provisions”. But even if the Chinese regulations are not yet as sophisticated as the EU’s AI Act, there are exceptions. Because one thing is clear: A large language model must not generate answers that contradict the party line.
While China’s regulations for AI remain general in many areas, the sub-legal standards for the Chinese AI industry already contain very fine-grained controls that go beyond the EU regulations. Among other things, they regulate how an AI should be trained, how it should be checked, how training data should be managed, and what qualifications people who annotate training data must have. This is why DeepSeek’s registration – which is required by law – contains the following sentence: “The AI only responds after illegal and negative information has been checked” 违法不良信息审核后. In Europe, only high-risk systems are checked so closely, including systems in critical infrastructure, security authorities, or personnel administration.
The verification of AI is often outsourced to service providers, and an industry has already emerged around the security verification of AI. In addition to private companies and semi-official think tanks, such as the China Academy of Information and Communications Technology (CAICT), universities also offer this service. And the creation of standards is not only taking place at state level. The commercial enterprise H3C, a subsidiary of Tsinghua University, was involved in the standards for the Chinese AI industry.
However, what constitutes safe AI is far from being conclusively clarified in China. There are several aspects, in addition to issues such as the threat posed to humans by AI, there are also ethical considerations that are the subject of lively debate. These include the question of whether people should have a right not to use AI. This is even supported by an AI Ethical Guideline from the Ministry of Science and Technology. But it is also about the security concerns of the authoritarian one-party regime, which wants to ensure that it remains in power.
And China is not alone in this. “China’s offer of safe AI can become an export hit,” says Daniel Sprick. “There are many jurisdictions or regimes in which an AI that is infused with Western values may not be the right thing, but rather an AI that is suitable for surviving an authoritarian security review.”
When Donald Trump signed a huge pile of documents in black marker pen on his first day in office, China watchers’ attention was primarily focused on the expected measures: new tariffs, further sanctions and a general tightening of trade policy. However, one crucial detail in Trump’s China plans initially went unnoticed.
In the memorandum, which deals with China’s trade practices, he not only ordered an investigation into unfair competition and further punitive tariffs. The investigation, the results of which are expected on April 1, will also explicitly assess China’s PNTR status.
The abbreviation PNTR stands for Permanent Normal Trade Relations and guarantees most countries in the world easy access to the US market. Withdrawing this status would make China an outcast trading partner, which is likely to further harden the political fronts. In fact, without PNTR, China would be in line with only four countries – Cuba, North Korea, Belarus and Russia – that are not currently in normal trade with the US.
As the news agency Reuters reports, a point has now been reached where the chances of the PNTR status being reversed in Washington are increasing. Calls to revoke China’s status have been coming from voices critical of China in the US Congress for years – from both Democrats and Republicans.
However, these opinions were never suitable for the majority. This could change if Trump also clearly backs the plans. Observers point out that the US president could also bring the trade status into play merely as a “nuclear option” in order to conclude a favorable deal with Beijing in the end. But no one can be sure. According to Reuters, some US companies are already making plans in case the status is revoked.
China was granted PNTR status in 2000, although the move was not uncontroversial even then, as the People’s Republic was far from meeting market economy criteria. However, the US government under Bill Clinton wanted to bring China into the World Trade Organization (WTO). Joining the WTO required stable trade relations with the USA. PNTR status was the prerequisite for this.
Despite criticism from trade unions and industry associations, which were already warning of job losses due to cheap Chinese imports, Clinton succeeded in getting the law through Congress in 2000. China was granted PNTR status, and the following year it was officially admitted to the WTO.
What followed was a historic economic boom that not only benefited China, but also US companies, which massively relocated production and at the same time opened up the Chinese market. However, some of the critics were also proved right: Since China’s WTO accession in 2001, trade volumes have multiplied, but the US has been running permanently high trade deficits. The flood of cheap Chinese imports brought US consumers lower prices, but also led to the closure of many American factories.
According to economists, turning back the clock now would have massive consequences for the US economy. The US system distinguishes between two main categories of tariffs. One category applies to countries with Most Favored Nation (MFN) status, i.e. most WTO members, including China under the current PNTR status. The second category applies to countries without trade privileges.
If China were to suddenly fall into the second category, it is estimated that average tariffs on Chinese goods could rise from the current level of around 19% to around 61%. This figure is suspiciously close to Trump’s election campaign threat to impose tariffs of 60 percent on China. This would effectively force China out of its market, but at a high economic price.
This drastic increase in tariffs would significantly increase the prices of Chinese products in the US, which could lead to rising costs for consumers and companies. An analysis by Oxford Economics predicts that the removal of PNTR status could reduce US gross domestic product by up to USD 1.9 trillion over a five-year period and cause the loss of around 801,000 jobs.
The blow to the Chinese economy could be even greater. An analysis by the Peterson Institute for International Economics from September 2024 predicts that China’s gross domestic product (GDP) could fall by around 0.6% as early as 2025 if PNTR status is revoked. In the long term, however, the effects would be more serious. There is talk of a “significant decline in GDP” over a longer period of time.
The institute warns that China’s attractiveness as a production location would be diminished. Companies could diversify their supply chains and relocate production facilities to other countries in order to avoid the increased tariffs. This would lead to a decline in foreign direct investment in China.
As expected, China itself reacted indignantly to the US debate. The considerations reflected a “Cold War mindset” that would throw trade back to earlier times. Withdrawing China’s PNTR status would be a clear violation of WTO rules. “It would harm the US itself and the global economy,” explains the Chinese Foreign Ministry.
The revocation of China’s PNTR status would also be a serious blow to the World Trade Organization (WTO) and its authority. The WTO is already under considerable pressure to maintain its relevance in an increasingly multipolar global economy. According to Christine McDaniel, a former US trade official, revoking China’s PNTR status would be a “big middle finger to the WTO”.
US plans to better control “AI diffusion” by introducing a comprehensive system of export quotas for the entire world have been met with great skepticism from other countries and the technology industry. But it seems that there is a consensus in Washington that this is the price to pay to prevent China from acquiring advanced AI for military purposes – and to secure US leadership by encouraging the development of a new AI infrastructure at home: Donald Trump has reversed several Biden-era executive orders, including one on AI safety, but regulations on the export of AI technologies have remained untouched.
The regulations divide the world into three “categories” – the US and 18 key partners in Category 1 are not subject to restrictions on the purchase of US-made high-end AI chips; China, Russia, and other countries subject to arms embargoes are still barred from purchasing these chips in Category 3; all other countries fall into the second category and must meet export quotas.
This is all reminiscent of the Cold War – the division of the world into blocs, and the upgrading of the status of AI to strategic weapons whose stockpiles must be controlled. The United States is taking these measures while it still holds most of the cards. Nvidia dominates the market for high-end chips, while its Chinese competitors are struggling to catch up. Chinese start-up DeepSeek may have shaken up the tech world by unveiling AI technology that uses less computing power – but the company is still reliant on the US giant’s advanced chips
The new regulations not only serve to block China’s access to AI, but also to strengthen the US position in this area. While the new administration still has time to make changes, the obvious compatibility of the regulations with Trump’s “America first” approach suggests that they will stick to this equally effective and clumsy tool. They give Washington control over all or significant transactions in AI chips and use caps to prevent other countries from reselling the limited supply to China. This approach represents a major shift in the way the US is trying to deny China access to AI – away from restrictions on specific companies and end uses to country-specific caps on AI purchases.
Trump’s first term initiated the “tech war” in 2018 by placing Chinese technology companies such as Huawei on the “Entity List”, which are subject to US export restrictions. And the Biden administration followed up in 2022 with China-wide restrictions on the use of certain chips and much stricter controls on “designated organizations”. However, this strategy was only partially effective, as the Chinese companies concerned found ways to circumvent the restrictions by disguising the intended end use or end user. This enabled them to source chips and chip manufacturing equipment via non-designated organizations and third countries – a loophole that is now to be closed with the export quotas for the 120 or so Category 2 countries.
The new regulations force Category 2 countries and companies to choose sides: Either live with the restrictions on American AI technology or join the United States and its allies in their fight against China’s technological ambitions. The countries and companies in this category that agree to comply with appropriate export controls and due diligence to prevent American AI technology from being diverted to China will in return receive a higher cap on the number of US chips. As with nuclear technology during the Cold War, the US is no longer willing to entrust its AI computing power to countries that do not play by the US rules.
Having previously banned the sale of chips and the equipment used to manufacture them, these new regulations are also designed to prevent China from gaining access to AI chips from abroad. Through foreign cloud services offering advanced AI, it was easy for Chinese companies to continue using this technology. Washington is now planning to tighten the requirements for existing “know-your-customer” regulations. These require US cloud providers to operate at least 50 percent of their computing power in the US and restrict access for untrusted countries. Foreign operators of data centers that participate in a new system for verifying end users and pass on information will in return be given the opportunity to purchase more modern US chips.
This extension of US extraterritorial restrictions has not gone down well with the European Union, as it divides the EU area into ten Category 1 countries, starting with Germany and France, which are exempt from any quotas, and 17 Category 2 countries. This poses challenges for the European single market and the geographical distribution of its data processing infrastructure. For example, a cloud provider serving customers in Category 2 country Austria may be tempted to locate its data centers in Germany, where it is not subject to any requirements. Even if the national quotas are set high enough and make little difference in the short term, the unilateral imposition of these rules has caused resentment in many European capitals.
The rules against “AI diffusion” prevent China from gaining access to advanced technology while ensuring that the US continues to set the tone on AI. Western allies also want to limit China’s military build-up – but few approve of Washington’s claim to AI supremacy. It is unlikely that countries around the world will be persuaded by Trump’s “America first” approach and Washington’s constant conflation of national security, economic security, and competitiveness. But until China (or any other country) can supply the world with its own high-end AI chips, they have no choice but to play by the US rules.
Antonia Hmaidi focuses on the geopolitics of technology, China’s pursuit of technological autonomy, particularly in areas such as semiconductors, operating systems and internet infrastructure, as well as China’s cyber security and hacking campaigns.
Wendy Chang’s research interests lie at the intersection of technology, politics and economics. Against the backdrop of increasing geopolitical tensions around the world, she explores how developments in areas such as artificial intelligence, self-driving cars, metaverse and quantum technologies affect China’s digital economy and its relations with other countries.
Editor’s note: Today more than ever, discussing China means debating controversially. We want to reflect the diversity of opinions so that you can gain an insight into the breadth of the debate. Opinion articles do not reflect the opinion of the editorial team.
Since the 1970s, the blue and white ice cream trucks of the Mister Softee brand have been part of Hong Kong’s street scene. The unmistakable melody of “On the Beautiful Blue Danube”, with which the trucks approach, evokes nostalgic childhood memories for many of the city’s residents. It was announced on Wednesday that Ho King-yuen, the man who brought the ice cream van brand to Hong Kong, died in Perth, Australia, at the age of 98. The former policeman had first seen the US brand’s mobile soft ice cream trucks on a trip to the UK. He was so enthusiastic about the idea that, together with two friends, he acquired the franchise rights for Mister Softee and brought the concept to Hong Kong. At the moment, 14 trucks are still driving through the city – for the surprise effect without fixed routes.
While Europe is forging ahead with strict AI rules, China is pursuing a pragmatic course – with strong ideological control. This model could be particularly attractive for governments in the Global South, writes Julia Fiedler. Because even though the EU has ten times as many well-intentioned regulations, China is far more determined in one area: censorship.
Donald Trump is toying with the idea of withdrawing China’s PNTR status. PNTR stands for Permanent Normal Trade Relations and guarantees most countries in the world easy access to the US market. Withdrawing the status would put China on a par with North Korea and Russia, explains Jörn Petring. The economic consequences would also be enormous for other countries.
AI export quotas divide the world into the good guys, the bad guys, and those who have to choose sides, write Antonia Hmaidi and Wendy Chang in today’s Opinion. They argue that the United States is trying to regulate technology while it still can.
Differing values, geopolitical competition, and the race for tech dominance: There are many obstacles to global AI governance that creates a framework for the ethical, security, and legal challenges posed by the technology. At the AI Action Summit in Paris earlier this week, 60 countries signed a final declaration calling for the initiation of a global dialog on AI governance, including Germany and China. The USA did not sign the declaration.
Who shapes the global regulations is crucial. China has already offered to take a leading role. In order to deal with AI together in the competition for influence, it can be helpful to understand how China’s AI legislation works, says Dr. Daniel Sprick, a research associate at the Institute for Chinese Legal Culture at the University of Cologne.
Standards play a central role here. These often primarily regulate aspects of connectivity or interoperability of technical products. However, China also uses standards as a form of sub-legal regulation to ensure that companies act in accordance with the rules. It is already able to export parts of its governance system that are shaped by its own authoritarian ideas.
Innovation needs freedom, but when it comes to artificial intelligence, there is great concern about misuse and harmful effects on society. So how do you draft a law for artificial intelligence that restricts AI without slowing down the breathless technological sprint? In the EU, the AI Act came into force in August 2024, the world’s first comprehensive law to regulate AI. While praised by some, it has also been criticized for its detailed and strict regulations. These could hinder innovation.
While the European regulation comprises almost 500 pages, the current Chinese legislation on AI does not even have 50 paragraphs in total. A comprehensive AI law 中华人民共和国人工智能法 is only being discussed in academic circles. This is despite the fact that regulation in China should actually be particularly strict – after all, even a large language model has to be true to the line. It is not allowed to provide answers on politically sensitive topics.
The fact that there is still no comprehensive AI law in China and that the few legal regulations are so meager may come as a surprise, but it is typical, says Sprick. Socialist law usually pours the program into a law first and then makes adjustments. Step by step, it is examined which further rules are useful – and these are then supplied in the form of standards.
The foundation stone was laid by the Social Credit System with its assessment standards. There are now numerous standards, some of which are also relevant for international consumers. For example, for cross-border e-commerce or data security.
In the field of AI, the Chinese government is clearly still unsure of where it wants to go. This can be seen from the fact that regulations are currently only laid down at the level of administrative guidelines and often have titles that begin with “provisional provisions”. But even if the Chinese regulations are not yet as sophisticated as the EU’s AI Act, there are exceptions. Because one thing is clear: A large language model must not generate answers that contradict the party line.
While China’s regulations for AI remain general in many areas, the sub-legal standards for the Chinese AI industry already contain very fine-grained controls that go beyond the EU regulations. Among other things, they regulate how an AI should be trained, how it should be checked, how training data should be managed, and what qualifications people who annotate training data must have. This is why DeepSeek’s registration – which is required by law – contains the following sentence: “The AI only responds after illegal and negative information has been checked” 违法不良信息审核后. In Europe, only high-risk systems are checked so closely, including systems in critical infrastructure, security authorities, or personnel administration.
The verification of AI is often outsourced to service providers, and an industry has already emerged around the security verification of AI. In addition to private companies and semi-official think tanks, such as the China Academy of Information and Communications Technology (CAICT), universities also offer this service. And the creation of standards is not only taking place at state level. The commercial enterprise H3C, a subsidiary of Tsinghua University, was involved in the standards for the Chinese AI industry.
However, what constitutes safe AI is far from being conclusively clarified in China. There are several aspects, in addition to issues such as the threat posed to humans by AI, there are also ethical considerations that are the subject of lively debate. These include the question of whether people should have a right not to use AI. This is even supported by an AI Ethical Guideline from the Ministry of Science and Technology. But it is also about the security concerns of the authoritarian one-party regime, which wants to ensure that it remains in power.
And China is not alone in this. “China’s offer of safe AI can become an export hit,” says Daniel Sprick. “There are many jurisdictions or regimes in which an AI that is infused with Western values may not be the right thing, but rather an AI that is suitable for surviving an authoritarian security review.”
When Donald Trump signed a huge pile of documents in black marker pen on his first day in office, China watchers’ attention was primarily focused on the expected measures: new tariffs, further sanctions and a general tightening of trade policy. However, one crucial detail in Trump’s China plans initially went unnoticed.
In the memorandum, which deals with China’s trade practices, he not only ordered an investigation into unfair competition and further punitive tariffs. The investigation, the results of which are expected on April 1, will also explicitly assess China’s PNTR status.
The abbreviation PNTR stands for Permanent Normal Trade Relations and guarantees most countries in the world easy access to the US market. Withdrawing this status would make China an outcast trading partner, which is likely to further harden the political fronts. In fact, without PNTR, China would be in line with only four countries – Cuba, North Korea, Belarus and Russia – that are not currently in normal trade with the US.
As the news agency Reuters reports, a point has now been reached where the chances of the PNTR status being reversed in Washington are increasing. Calls to revoke China’s status have been coming from voices critical of China in the US Congress for years – from both Democrats and Republicans.
However, these opinions were never suitable for the majority. This could change if Trump also clearly backs the plans. Observers point out that the US president could also bring the trade status into play merely as a “nuclear option” in order to conclude a favorable deal with Beijing in the end. But no one can be sure. According to Reuters, some US companies are already making plans in case the status is revoked.
China was granted PNTR status in 2000, although the move was not uncontroversial even then, as the People’s Republic was far from meeting market economy criteria. However, the US government under Bill Clinton wanted to bring China into the World Trade Organization (WTO). Joining the WTO required stable trade relations with the USA. PNTR status was the prerequisite for this.
Despite criticism from trade unions and industry associations, which were already warning of job losses due to cheap Chinese imports, Clinton succeeded in getting the law through Congress in 2000. China was granted PNTR status, and the following year it was officially admitted to the WTO.
What followed was a historic economic boom that not only benefited China, but also US companies, which massively relocated production and at the same time opened up the Chinese market. However, some of the critics were also proved right: Since China’s WTO accession in 2001, trade volumes have multiplied, but the US has been running permanently high trade deficits. The flood of cheap Chinese imports brought US consumers lower prices, but also led to the closure of many American factories.
According to economists, turning back the clock now would have massive consequences for the US economy. The US system distinguishes between two main categories of tariffs. One category applies to countries with Most Favored Nation (MFN) status, i.e. most WTO members, including China under the current PNTR status. The second category applies to countries without trade privileges.
If China were to suddenly fall into the second category, it is estimated that average tariffs on Chinese goods could rise from the current level of around 19% to around 61%. This figure is suspiciously close to Trump’s election campaign threat to impose tariffs of 60 percent on China. This would effectively force China out of its market, but at a high economic price.
This drastic increase in tariffs would significantly increase the prices of Chinese products in the US, which could lead to rising costs for consumers and companies. An analysis by Oxford Economics predicts that the removal of PNTR status could reduce US gross domestic product by up to USD 1.9 trillion over a five-year period and cause the loss of around 801,000 jobs.
The blow to the Chinese economy could be even greater. An analysis by the Peterson Institute for International Economics from September 2024 predicts that China’s gross domestic product (GDP) could fall by around 0.6% as early as 2025 if PNTR status is revoked. In the long term, however, the effects would be more serious. There is talk of a “significant decline in GDP” over a longer period of time.
The institute warns that China’s attractiveness as a production location would be diminished. Companies could diversify their supply chains and relocate production facilities to other countries in order to avoid the increased tariffs. This would lead to a decline in foreign direct investment in China.
As expected, China itself reacted indignantly to the US debate. The considerations reflected a “Cold War mindset” that would throw trade back to earlier times. Withdrawing China’s PNTR status would be a clear violation of WTO rules. “It would harm the US itself and the global economy,” explains the Chinese Foreign Ministry.
The revocation of China’s PNTR status would also be a serious blow to the World Trade Organization (WTO) and its authority. The WTO is already under considerable pressure to maintain its relevance in an increasingly multipolar global economy. According to Christine McDaniel, a former US trade official, revoking China’s PNTR status would be a “big middle finger to the WTO”.
US plans to better control “AI diffusion” by introducing a comprehensive system of export quotas for the entire world have been met with great skepticism from other countries and the technology industry. But it seems that there is a consensus in Washington that this is the price to pay to prevent China from acquiring advanced AI for military purposes – and to secure US leadership by encouraging the development of a new AI infrastructure at home: Donald Trump has reversed several Biden-era executive orders, including one on AI safety, but regulations on the export of AI technologies have remained untouched.
The regulations divide the world into three “categories” – the US and 18 key partners in Category 1 are not subject to restrictions on the purchase of US-made high-end AI chips; China, Russia, and other countries subject to arms embargoes are still barred from purchasing these chips in Category 3; all other countries fall into the second category and must meet export quotas.
This is all reminiscent of the Cold War – the division of the world into blocs, and the upgrading of the status of AI to strategic weapons whose stockpiles must be controlled. The United States is taking these measures while it still holds most of the cards. Nvidia dominates the market for high-end chips, while its Chinese competitors are struggling to catch up. Chinese start-up DeepSeek may have shaken up the tech world by unveiling AI technology that uses less computing power – but the company is still reliant on the US giant’s advanced chips
The new regulations not only serve to block China’s access to AI, but also to strengthen the US position in this area. While the new administration still has time to make changes, the obvious compatibility of the regulations with Trump’s “America first” approach suggests that they will stick to this equally effective and clumsy tool. They give Washington control over all or significant transactions in AI chips and use caps to prevent other countries from reselling the limited supply to China. This approach represents a major shift in the way the US is trying to deny China access to AI – away from restrictions on specific companies and end uses to country-specific caps on AI purchases.
Trump’s first term initiated the “tech war” in 2018 by placing Chinese technology companies such as Huawei on the “Entity List”, which are subject to US export restrictions. And the Biden administration followed up in 2022 with China-wide restrictions on the use of certain chips and much stricter controls on “designated organizations”. However, this strategy was only partially effective, as the Chinese companies concerned found ways to circumvent the restrictions by disguising the intended end use or end user. This enabled them to source chips and chip manufacturing equipment via non-designated organizations and third countries – a loophole that is now to be closed with the export quotas for the 120 or so Category 2 countries.
The new regulations force Category 2 countries and companies to choose sides: Either live with the restrictions on American AI technology or join the United States and its allies in their fight against China’s technological ambitions. The countries and companies in this category that agree to comply with appropriate export controls and due diligence to prevent American AI technology from being diverted to China will in return receive a higher cap on the number of US chips. As with nuclear technology during the Cold War, the US is no longer willing to entrust its AI computing power to countries that do not play by the US rules.
Having previously banned the sale of chips and the equipment used to manufacture them, these new regulations are also designed to prevent China from gaining access to AI chips from abroad. Through foreign cloud services offering advanced AI, it was easy for Chinese companies to continue using this technology. Washington is now planning to tighten the requirements for existing “know-your-customer” regulations. These require US cloud providers to operate at least 50 percent of their computing power in the US and restrict access for untrusted countries. Foreign operators of data centers that participate in a new system for verifying end users and pass on information will in return be given the opportunity to purchase more modern US chips.
This extension of US extraterritorial restrictions has not gone down well with the European Union, as it divides the EU area into ten Category 1 countries, starting with Germany and France, which are exempt from any quotas, and 17 Category 2 countries. This poses challenges for the European single market and the geographical distribution of its data processing infrastructure. For example, a cloud provider serving customers in Category 2 country Austria may be tempted to locate its data centers in Germany, where it is not subject to any requirements. Even if the national quotas are set high enough and make little difference in the short term, the unilateral imposition of these rules has caused resentment in many European capitals.
The rules against “AI diffusion” prevent China from gaining access to advanced technology while ensuring that the US continues to set the tone on AI. Western allies also want to limit China’s military build-up – but few approve of Washington’s claim to AI supremacy. It is unlikely that countries around the world will be persuaded by Trump’s “America first” approach and Washington’s constant conflation of national security, economic security, and competitiveness. But until China (or any other country) can supply the world with its own high-end AI chips, they have no choice but to play by the US rules.
Antonia Hmaidi focuses on the geopolitics of technology, China’s pursuit of technological autonomy, particularly in areas such as semiconductors, operating systems and internet infrastructure, as well as China’s cyber security and hacking campaigns.
Wendy Chang’s research interests lie at the intersection of technology, politics and economics. Against the backdrop of increasing geopolitical tensions around the world, she explores how developments in areas such as artificial intelligence, self-driving cars, metaverse and quantum technologies affect China’s digital economy and its relations with other countries.
Editor’s note: Today more than ever, discussing China means debating controversially. We want to reflect the diversity of opinions so that you can gain an insight into the breadth of the debate. Opinion articles do not reflect the opinion of the editorial team.
Since the 1970s, the blue and white ice cream trucks of the Mister Softee brand have been part of Hong Kong’s street scene. The unmistakable melody of “On the Beautiful Blue Danube”, with which the trucks approach, evokes nostalgic childhood memories for many of the city’s residents. It was announced on Wednesday that Ho King-yuen, the man who brought the ice cream van brand to Hong Kong, died in Perth, Australia, at the age of 98. The former policeman had first seen the US brand’s mobile soft ice cream trucks on a trip to the UK. He was so enthusiastic about the idea that, together with two friends, he acquired the franchise rights for Mister Softee and brought the concept to Hong Kong. At the moment, 14 trucks are still driving through the city – for the surprise effect without fixed routes.
