• AiNews.com
  • Posts
  • Chinese Entities Use Amazon Cloud to Access Banned U.S. AI Technology

Chinese Entities Use Amazon Cloud to Access Banned U.S. AI Technology

An image illustrating Chinese entities accessing U.S. AI technology and chips through cloud services. The image features cloud servers and digital interfaces displaying AI chips like Nvidia A100 and H100, with connections to the U.S. and Chinese flags via digital pathways. Visual cues such as locks and barriers symbolize export restrictions, reflecting the ongoing geopolitical and technological tensions

Image Source: ChatGPT

Chinese Entities Use Amazon Cloud to Access Banned U.S. AI Technology

Recent documents reveal that state-linked Chinese entities have been using Amazon Web Services (AWS) and other cloud providers to access advanced U.S. chips and artificial intelligence (AI) capabilities that are otherwise restricted by U.S. export regulations. This move highlights the ongoing challenges in enforcing technology embargoes and the creative strategies being employed to circumvent them.

Accessing Restricted U.S. Technology Through the Cloud

Over the past two years, the U.S. government has imposed stringent restrictions on the export of high-end AI chips to China, aiming to curb the Chinese military's capabilities. However, these restrictions primarily target the physical export or transfer of technology, leaving a loophole that allows access through cloud services.

A review of more than 50 public tender documents, conducted by Reuters, showed that at least 11 Chinese entities have sought to bypass these restrictions by accessing U.S. technologies or cloud services indirectly. Four of these entities specifically named AWS as their cloud service provider, although they accessed the services through Chinese intermediaries rather than directly from AWS.

For example, in March 2024, Shenzhen University spent approximately $28,000 on an AWS account to access cloud servers powered by Nvidia A100 and H100 chips, which are critical for training large-language models like OpenAI's ChatGPT. Similarly, Zhejiang Lab, a research institute working on its own AI model, GeoGPT, planned to spend around $26,000 on AWS cloud computing services after struggling to obtain sufficient computing power from China's Alibaba Cloud.

Chinese Demand for U.S. AI Capabilities

The demand for U.S. AI chips and cloud computing power is driven by China's ambitions to advance its AI capabilities. Nvidia’s A100 and H100 chips, which are at the center of U.S. export restrictions, are highly sought after for their ability to power sophisticated AI models. Although the U.S. banned the export of these chips to China in September 2022, Chinese entities have found ways to access them through cloud services.

This method of accessing restricted technology does not currently violate U.S. regulations, which only govern the direct export or transfer of commodities and software. However, the U.S. government is aware of this loophole and is taking steps to close it. In April 2024, legislation was introduced in Congress to empower the Commerce Department to regulate remote access to U.S. technology through the cloud.

Efforts to Tighten Restrictions

U.S. lawmakers and officials are increasingly concerned about China’s ability to bypass export controls through cloud computing. Michael McCaul, chair of the U.S. House of Representatives Foreign Affairs Committee, has expressed the need to address this loophole, which he says has been a concern for years.

The Commerce Department has proposed new rules that would require U.S. cloud computing services to verify users of large AI models and report their activities to regulators. These rules, if finalized, could also grant the Commerce Secretary the authority to impose prohibitions on specific customers.

Microsoft and Other Cloud Providers

Amazon is not the only cloud provider targeted by Chinese entities. Sichuan University, for instance, has sought to purchase 40 million Microsoft Azure OpenAI tokens for its generative AI platform, according to a tender document from April 2024. The university’s procurement was fulfilled by a Chinese intermediary, further illustrating the strategies being used to access restricted U.S. technology.

The revelations of these practices underscore the complexity of enforcing technology embargoes in a globalized and highly interconnected digital landscape. As the U.S. government seeks to tighten its regulations, the ongoing race to develop and deploy advanced AI models continues to escalate.