Tech Industry Calls for Changes to AI Diffusion Rule, Tier Structure

The U.S. export control strategy “should do more than restrict the flow of AI technologies” to China -- it should “ensure that America is ‘winning diffusion,’” Open AI said in public comments to the Office of Science and Technology Policy in response to the Trump administration's AI “action plan” (see 2502070065). The company said that even though America has a “lead on AI today,” the recent success of China’s AI model DeepSeek (see 2502030031 and 2501300067) “shows that our lead is not wide and is narrowing,” it said.

The diffusion rule, released in January, places global license requirements on exports of advanced AI chips, introduces a new framework for preapproved data facilities under the Bureau of Industry and Security's validated end-user program, creates new controls on certain closed artificial intelligence model weights, and more (see 2501130026).

The restrictions have been widely panned by technology industries both in the U.S. and abroad (see 2502270074, 2502110074 and 2503110027), partly because they subject many close American trading partners, including some EU member states, to new AI-related license requirements. Some EU members are in the rule’s Tier 1 category, the group of nations that will face the fewest export restrictions, while others are in the more restrictive Tier 2.

EU officials said last month they were planning to push Trump officials to lift the restrictions or move all EU members to Tier 1 (see 2502200051). Digital Europe, a trade policy group, said EU nations are being treated “unequally” by the rule, and this has “created tensions.”

“Rather than introducing restrictive measures, the focus should be on fostering collaboration on mutual priorities,” the group said in public comments. “Both the EU and the US depend on one another for security and to remain competitive. This interdependence should be seen as a strength.”

OpenAI said the Trump administration should change the AI diffusion rule to include more countries in Tier 1 and limit Tier 2 to countries with a “history of failing to prevent export-controlled chips and other US-developed IP” from being used by Tier 3 countries. Tier 3 countries would still include China and a “small cohort of countries aligned” with China’s government, OpenAI said.

The U.S. should also create a “transparent process to evaluate countries’ readiness to transition from” Tier 2 to Tier 1 and help Tier 2 governments “strengthen their in-country security programs.”

It should also allow exports of advanced AI chips to an end-user in a Tier II country as long as that end-user “meets Tier I security requirements” and “puts in place additional corporate governance controls as well as technology-enhanced protections” to protect against diversion. OpenAI especially mentioned “hardware-enabled mechanisms,” which could be a reference to location-tracking features or other hardware in a chip that would allow it to be used only in situations that comply with export controls or the terms of an export license (see 2401080060).

Google didn’t say whether it supported the rule’s existing tier structure, but it said U.S. export controls should be “carefully crafted to support legitimate market access for U.S. businesses while targeting the most pertinent risks.” The existing AI diffusion rule “may undermine economic competitiveness goals the current Administration has set by imposing disproportionate burdens on U.S. cloud service providers,” the company added.

It said it plans to submit a “more detailed analysis of the AI Diffusion rule” before the May 15 comment deadline.

Google also said BIS may need more resources to enforce export controls, including its own set of “cutting-edge AI tools for supply chain monitoring and counter-smuggling efforts.” The agency should also “streamline” its export licensing processes and consider rules that go “beyond limits on hardware exports.”

The Software & Information Industry Association said it supports strong AI chip controls against China and other “adversaries,” but it also cautioned against broad restrictions on AI software, cloud services or trained AI models, “which would be difficult to enforce and could handicap U.S. industry more than our rivals.” It said existing U.S. proposals to place export controls on cloud computing services or AI model access would “pose unworkable implementation challenges and will give China a leg up in the global AI competition.”

The R Street Institute, a center-right Washington think tank, said it’s concerned the U.S. is focusing too heavily on export controls and not enough on boosting American innovation. The institute said the U.S. won’t win its technology competition against China by only imposing restrictions, pointing to the success of China’s DeepSeek.

“Even with various export controls and other domestic computing constraints in place, Chinese AI developers have made major technological advances -- especially with open-source capabilities,” the R Street Institute said. “The key to maintaining global AI leadership is to remain on the cutting edge of the technological frontier.”