US tackles potential Chinese AI threat

For months, officials in the Biden administration have been preparing a new executive order to limit U.S. investment in one geopolitical rival: China. Their goal is to block US private equity and venture capital from contributing to the development of cutting-edge technologies that could aid the Chinese military.

America’s efforts to maintain its technological superiority over China, including a ban on the export of some advanced semiconductors last year, have become a defining issue in relations with China. Secretary of State Antony Brinken is scheduled to visit Beijing this weekend, where he is likely to face complaints from Chinese officials who see moves such as investment restrictions as an attempt by the United States to curb China’s economic growth. .

U.S. officials, meanwhile, have argued that upcoming capital controls are targeted measures against what they perceive as national security threats. They have sought to develop investment rules that do not jeopardize the broader flows of investment and trade between the world’s two largest economies.

But distinguishing between technology that the Chinese government can use to boost its military power and technology that Chinese companies use for everyday commercial purposes will be difficult for the Biden administration, especially when it comes to AI, according to people familiar with the deliberations. It turned out that

The conundrum is one of the outstanding issues with an executive order expected to ban investments in advanced semiconductors and quantum computing as part of a year-long pilot program, the people said. Some forms of AI investment may be prohibited under the rules or simply subject to new disclosure requirements, the people said.

“AI is a nonsensical category in many ways, including everything from Netflix recommendation algorithms to autonomous weapon systems and many things in between,” says the Peterson International Economics Study. “It’s very difficult to define,” said Martin Chozempa, who studies capital and technology management at the Institute.

A spokesman for the National Security Council said the United States was making progress on an executive order that investors and lobbyists had been waiting for months.

“It’s a complicated process and we want to make sure we get it right, but it will take time,” the spokesperson said.

While some forms of AI are developed to perform specific functions, many AI companies are focused on building general-purpose systems that can be trained to perform all kinds of tasks.

Global investment in AI startups is booming thanks to rapid consumer adoption of ChatGPT and other so-called generative AI tools that can instantly create text, images, videos and computer code based on written prompts. is presenting.

This technology is expected to find wide commercial application. But AI models designed for computer coding could easily be used for hacking, while models intended to help create medicines could also produce new chemical weapons. , said the researchers.

“If you’re using AI to generate images, it could potentially be used to run war simulations or used in games,” said Daniel Castro, vice president of the think tank Information Technology Innovation Foundation. .

Other forms of AI can pose similar challenges. US mobile phone chip maker Qualcomm’s venture capital arm is based in Beijing to develop an AI-powered tool that tracks eye movements in 2016, according to researchers at Georgetown University in a recent report. invested in the start-up 7Invensun, which puts US venture company invests in Chinese AI.

7Invensun’s technology could be used in virtual reality goggles and other consumer products, but it could also have military or security applications such as infrared facial recognition cameras and fighter pilot training. Georgetown researchers Emily Weinstein and Ngor Luong wrote that the company has revealed that it is working with the Chinese state-owned defense enterprise and the China Air Force Aviation University.

Representatives for Qualcomm and 7Invensan did not respond to requests for comment.

A policy that bans investments like Qualcomm’s would be difficult to articulate and enforce. Examining the code of a particular AI system does not always reveal its capabilities, as the system can be rapidly trained on new data to accomplish new tasks. Advanced AI systems can involve highly complex computing processes that produce unexpected results, making it difficult to put limits on their output.

“It’s very difficult to make sure these systems are secure. They’re essentially black boxes,” said Tim Fist, a researcher at the Center for a New American Security.

According to a Georgetown report, US investors were involved in 401 deals with Chinese AI companies from 2015 to 2021, with US investors alone investing 7.4 billion over the period. reached $50 million.

But fears of curbing US investment in China’s cutting-edge technology have already pushed some US companies out of the market. Renowned venture capital firm Sequoia recently spun off its China operations amid tensions between Washington and China. According to the Institute of International Finance, overall net foreign direct investment in China will reach its lowest level in nearly two decades in 2023.

A spokeswoman for the Chinese Embassy in Washington said the United States and China should pursue healthy economic competition.

“Discriminatory regulations targeting companies of certain nationalities run counter to the basic principles of international economics and trade,” the spokesman said. I will defend it resolutely,” he said.

The Biden administration’s previous ban on advanced semiconductor exports would be one way to prevent AI companies from accessing the computing power needed to develop the most sophisticated models. The export ban has prompted some Chinese AI companies to try to develop advanced AI without cutting-edge chips. The US has previously banned the export of AI technology used to automate geospatial imagery.

Still, Biden administration officials worry that U.S. investors could transfer valuable knowledge and expertise to Chinese startups, allowing them to develop their own versions of advanced technologies, including semiconductors. there is U.S. venture capital firms often provide investee companies with access to industry knowledge and connections that may not be available elsewhere.

“Investing in China is never completely risk-free,” said Weinstein, a fellow at the Center for Security and Emerging Technologies. “The administration needs to decide how much. It’s about being able to take risks,” he said. and draw a line there. ”

Email Andrew Duehren (andrew.duehren@wsj.com) and Ryan Tracy (ryan.tracy@wsj.com).