Chinese firms use AWS to access restricted US AI chips

Chinese firms are reportedly gaining access to cutting-edge AI capabilities and restricted US Chips through Amazon cloud services or its competitors, according to details obtained from some tender documents.

This comes as the US government implemented embargoes on the export of AI chips and tools due to geopolitical tensions between the US and China. According to the US, a ban on exports of high-end AI chips to China is to limit the Chinese military’s abilities.

Amazon is not breaking the law

Tender documents seen by Reuters portrayed the extent to which Chinese companies have been tactful in accessing advanced computing power and access generative AI models. One such tactic they have used is through the Amazon Web Services (AWS).

For Amazon, providing those chips through cloud services is not in violation of the current laws as they are only explicit on export or transfer of products, software, or technology.

Reuters revealed they reviewed over 50 papers in the past year obtained from an open Chinese database. The documents revealed that over 11 Chinese companies gained access to restricted US technologies or cloud amenities.

From these, four companies clearly indicated their involvement with Amazon, stating that they access the facilities through Chinese mediators and not directly from AWS.

“AWS complies with all applicable US laws including trade laws, regarding the provision of AWS services inside and outside of China.”

Amazon’s spokesperson.

The papers also show that US companies are cashing in on the increasing demand for computing power from China.

Canalys, a research firm, argues that AWS constantly reins close to a third of the universal infrastructure market. Another research firm, IDC, notes that AWS occupies the sixth position in China when it comes to cloud service provision.

Chinese firms not getting enough from local suppliers

A March tender document reveals that Shenzhen University used up over 200,000 Yuan or $27,996 on an AWS account to access cloud servers which were driven by Nvidia A100 and H100 chips.

The documents show that the tertiary institution had access through a middleman, Yunda Technology Ltd Co. The US put restrictions on the trade of two Nvidia chips, used to power large language models (LLM), like OpenAI’s ChatGPT.

Yunda Technology and Shenzhen University did not respond to questions sent to them while Nivdia failed to comment on Shenzhen University’s spending or any of the Chinese companies’ transactions.

According to Reuters’ investigation, a research institute, Zhejiang Lab which is developing its own LLM (GeoGPT), revealed in an April paper that it was looking at spending above 184,000 Yuan to acquire AWS cloud computing services as its AI models could not get adequate computing power from local supplier, Alibaba.

However, a Zhejiang Lab representative said they did not complete the acquisition and did not respond to questions regarding the logic behind their choice or how they met the processing power criteria of their LLM.025

US worries about the cloud computing loopholes

The US government is currently working to strengthen restrictions to also limit cloud access. Michael McCaul, the US House of Representatives Foreign Affairs Committee chair, said in a statement that they were concerned about foreign entities accessing advanced US computing capabilities through the cloud.

“This loophole has been a concern of mine for years and we are long overdue to address it.”

– McCaul.

In April, the Congress Department was permitted to introduce laws that control the remote access of US technology. However, it remains uncertain whether the law was passed.

“We are working closely with Congress and seeking additional resources to strengthen our existing controls that restrict PRC companies from accessing advanced AI chips through remote access to cloud computing capability,” said a department spokesperson

In January, the Commerce Department also pushed for a law that would require US cloud computing services to authenticate large AI model users and report to regulators when using US cloud computing to train massive AI models capable of “malicious cyber-enabled activity.”

While the law is yet to be finalized, the commerce secretary is expected to be able to enforce an embargo on customers.

“We are aware the commerce department is considering new regulations, and we comply with all applicable laws in the countries in which we operate,” the AWS spokesperson said.

Demand for cloud services by Chinese firms remains strong

According to a tender document by the University of Science and Technology of China’s (USTC) Suzhou Institute of Advanced Research, the institute wanted to rent 500 cloud servers, each of these powered by Nvidia’s A100 chips.

The Chinese firms are also looking to access Microsoft’s cloud services. Sichuan University said in an April tender document that it was building a generative AI platform and purchasing 40 million Microsoft Azure OpenAI tokens to support this project.

According to Reuters, the university’s procurement document in May showed that the tokens were supplied by Sichuan Province Xuedong Technology Co Ltd.OpenAI said its services are not available in China, adding that Azure OpenAI operates under Microsoft’s policies.

Source: https://www.cryptopolitan.com/chinese-firms-use-aws-to-access-us-ai-chips/