The Biden-Harris Administration has introduced a new Interim Final Rule on Artificial Intelligence Diffusion aimed at enhancing US national security and preventing the misuse of advanced US technology by countries of concern.
The rule strengthens protections against misuse of advanced AI technologies by countries of concern.
Measures defined in the new rule include:
-
Continuing restrictions on advanced semiconductor sales to prevent their use in training sophisticated AI systems
-
Limiting the transfer of model weights for advanced closed-weight models
-
Setting security protocols to protect closed-weight AI models
Licensing for Chip Sales
The rule further streamlines licensing for chip sales to ensure responsible global diffusion of AI technology. It removes restrictions on chip sales to 18 key allies and partners with strong technology protection frameworks, facilitating easier access for jurisdictions aligned with US security interests.
Chip orders involving computing power up to approximately 1700 advanced GPUs will not require a license. This exemption covers most chip orders from universities, medical institutions and research organizations using AI for non-sensitive purposes. The measure also accelerates the shipment of low-risk US technology worldwide.
New High Security Standards
To ensure the security of AI technology, the rule introduces two key statuses for entities meeting high security and trust standards:
-
Universal Verified End User (UVEU): Granted to entities in allied nations, allowing them to place up to 7% of their global AI computational capacity in other countries, facilitating expansion while keeping frontier training within the US
-
National Verified End User (NVEU): Available to entities outside countries of concern, permitting purchases of computational power equivalent to up to 320,000 advanced GPUs over two years
Entities not classified under these categories can still acquire up to 50,000 advanced GPUs per country, ensuring US technology supports sectors like healthcare and telecommunications worldwide.
AI as a “Great Power” Technology
Industry experts have shared mixed reactions to the new rule, highlighting both its strengths and potential downsides. Casey Ellis, founder of Bugcrowd, noted that the rule reflects a “broader consensus in Washington that AI is establishing itself as a ‘Great Power’ technology,” emphasizing the importance of maintaining US and allied dominance while cautioning against overly restrictive export controls that could stifle innovation.
Kris Bondi, CEO and co-founder of Mimoto, expressed concern that the rule’s approach might limit global collaboration, stating, “While I agree that the use and protection of AI are critical for US national security and economic strength, this form of isolationism will undermine innovation.”
Stephen Kowski, Field CTO at SlashNext, acknowledged the need for strict controls but stressed that they should be carefully targeted. “The focus should be on creating precise controls that protect critical AI capabilities while fostering collaboration with trusted partners,” Kowski explained.
These steps build on previous export controls from October 2022 and 2023 aimed at safeguarding US national security while maintaining international collaboration on secure AI development.
Image credit: rarrarorro / Shutterstock.com