(NewsNation) — Chinese organizations are using cloud services to access the microchips that power artificial intelligence (AI) tools in the United States, Reuters reported.
At least seven Chinese entities have tried to access restricted cloud services and other technology in the U.S. in the past year, according to Reuters’ analysis of documents on public Chinese databases. The findings shed light on an information loophole at a time when the competition is heating up over the nations’ AI capabilities.
The U.S. has export restrictions in place to prevent Chinese firms from buying the chips American companies use to run technology such as ChatGPT. However, U.S. laws only oversee the direct transfer of physical technology and software. That means it’s technically not illegal for Chinese firms to access American-made chips and AI tools via cloud services, according to a report from the tech and IT magazine CIO.
Legislators introduced a bill earlier this year to close that loophole. In the meantime, Chinese firms have been seeking access to U.S. AI capabilities through restricted cloud services from giants like Amazon and Microsoft, Reuters reported.
“(Amazon Web Services) complies with all applicable U.S. laws, including trade laws, regarding the provision of AWS services inside and outside of China,” a spokesperson for Amazon’s cloud business told Reuters.
The news outlet reported that Shenzhen University spent 200,000 yuan ($27,996) on an Amazon Web Services (AWS) account to access cloud servers powered by Nvidia A100 and H100 machine learning chips for an unspecified project. The university reportedly accessed the service through an intermediary called Yunda Technology Ltd Co.
Other Chinese entities seeking access to AI models and computing power via AWS include the state-run research institute Zhejiang Lab and the National Center of Technology Innovation for EDA. The latter is a state-backed agency that helps Chinese companies develop blueprints for chip designs. Several colleges, universities, research institutes and one car maker were also on Reuters’ list of entities that tried to access restricted U.S. technology through the cloud.
China and the U.S. both are ramping up efforts to restrict access to data and strengthen control over AI. But as the technology rapidly develops, restrictions surrounding proprietary AI models have been a work in progress.
Chinese authorities, including President Xi Jinping, have pushed for engineers to develop original and “controllable” AI technology.
However, China has relied on Western open-source models such as Meta’s Llama. Earlier this year, the Beijing Academy of Artificial Intelligence said most Chinese AI models were built using Llama models, according to a separate Reuters report.
Reuters contributed to this report.
