U.S.' AI export curbs dubbed unworkable

Measures go against Washington's goal of maintaining leadership, experts say

Experts have sharply criticized the latest export controls on artificial intelligence models and chips implemented by outgoing U.S. President Joe Biden's administration, describing the measures as unworkable and counterproductive to the U.S.' stated goals of protecting national security and maintaining AI leadership.

The administration unveiled its Interim Final Rule for an "Export Control Framework for Artificial Intelligence Diffusion" last week as it approaches the end of its term.

While the framework aims to safeguard U.S. national security by limiting the export of advanced AI chips and technologies to strategic competitors, particularly China, it has sparked intense debate about its potential impact on U.S. competitiveness in the global market.

The new regulations, which took effect on Jan 13, include a grace period for compliance until May 15. The framework establishes a tiered system of export controls, with only 18 "trusted" economies designated as "tier one" and exempted from strict export caps. This select group includes close U.S. allies such as the United Kingdom, Japan, and Germany.

The framework's "tier two" designation encompasses most of the world, including close U.S. partners like Israel, Singapore, Brazil, India, Saudi Arabia, and the United Arab Emirates. These nations face substantial restrictions with stringent country-specific limits on chip imports. China falls into the most restrictive "tier three" category which the U.S. considers "adversaries".

The regulations particularly affect the export of graphics processing unit chips essential for powering cutting-edge AI applications, including large language models. Industry experts warn that the framework's broad scope and lack of targeted measures could undermine U.S. competitiveness in AI, weaken U.S. global leadership in digital policy, and create opportunities for foreign AI chip manufacturers to capture greater market share.

"It's a nice, cute touch, but it doesn't actually work, and we know that from experience," said James Andrew Lewis, senior vice-president and director of the Strategic Technologies Program at the Center for Strategic and International Studies.

Speaking at a webinar organized by the Information Technology and Innovation Foundation on Thursday, Lewis said if the goal is to impede China's AI development, the framework will not work.

"It's worthwhile to accelerate what the U.S. is doing. It's worthwhile to seek to dominate the market, but it's less useful to try and stop the Chinese. Because 10 years ago the Chinese were still dependent on the U.S. for technology, and that's increasingly less true, they'll come up with workarounds. They'll come up with substitutes," he said.

Objectives questioned

Questioning the policy's fundamental objectives, Lewis suggested that instead of trying to constrain China, the focus should be on "having the biggest market share for American companies possible".

"When you design a regulation, you should have some goals in mind, and the goal we should have in mind for this regulation is how to accelerate U.S. leadership in this market, and the best measure for that is market share," Lewis said.

Ken Glueck, executive vice-president of Oracle, shared these concerns, arguing the new rules would foster new foreign competition for U.S. companies and describing the "arbitrary market allocation scheme" as "just unworkable".

"The problem is that there are far too many tier two countries," Glueck said at the webinar. From a "practitioner's view", he said, "You are not allowed to deploy more than 50 percent of GPUs outside of the United States, that means U.S. industry is a net loser under this discussion."

"If tariffs can be considered the first phase of America's economic war with China, then blacklisting appears to be the second phase. Of course, with tariffs, a Chinese company could still do business with the U.S., but those that are blacklisted are denied such opportunities," Anthony Moretti, department head and an associate professor of the Communication and Organizational Leadership Department at Robert Morris University in Pennsylvania, told China Daily.

"Biden has become even more of a China hawk since his vice-president lost the 2024 election; there have been multiple blacklist announcements over the past eight to 10 weeks, and they raise questions about just how sincere he was in supporting the San Francisco vision and any other efforts, for that matter, that might have improved U.S.-China relations," Moretti said.

"Regardless of the reason or reasons, the erosion of the bilateral relationship continues, and that is not good for either country or for the global community as a whole," he said.