The Biden-Harris Administration has unveiled new guidelines designed to regulate chip licensing and enhance artificial intelligence (AI) security standards. Released on Monday, the Interim Final Rule on Artificial Intelligence Diffusion builds on existing measures to protect U.S. technological leadership while addressing national security concerns.
Strengthening AI Security and Addressing Smuggling:
The new rule aims to prevent the smuggling of sensitive technologies and close existing loopholes that could undermine U.S. national security. It seeks to ensure that the U.S. maintains control over key technologies, with the White House emphasizing that AI systems and chips should “run on American rails” to safeguard both national security and economic strength.
Key Aspects of the New Guidelines:
The rule introduces several critical measures for regulating the global distribution of U.S. technology:
- Chip Sales Restrictions: Sales of semiconductors to 18 key allies and partners will be streamlined to facilitate large-scale purchases. The rule eliminates licensing requirements for chip orders involving computational power equivalent to approximately 1,700 advanced GPUs.
- Verified End User Status: Entities that meet strict security standards can obtain special designations, including Universal Verified End User (UVEU) status, which grants access to trusted technology. This status is available to entities in close allied nations. Similarly, the National Verified End User (NVUE) status allows entities from non-concerned countries to purchase computational power equivalent to up to 320,000 GPUs over the next two years.
- Non-VEU Entities: Non-verified entities in non-ally nations can still access large amounts of computational power, though with a cap of 50,000 advanced GPUs per country, ensuring broader access to technology for businesses and government services.
- International Collaboration: Governments are encouraged to establish government-to-government agreements to promote international cooperation, fostering shared values in AI development, deployment, and security.
Protecting U.S. Technology from Adversaries:
The rule introduces stricter measures to prevent “countries of concern,” such as China, from accessing critical AI technologies. This includes ensuring advanced semiconductors cannot be used by these countries to train cutting-edge AI models. It also restricts the transfer of model weights for advanced AI systems to untrusted parties and sets new security standards to protect these models.
Context and Background:
The new guidelines follow previous actions by the U.S. government, including an August 2023 executive order that limits investments in Chinese tech companies related to sensitive technologies. In 2022, the U.S. imposed restrictions on AI-focused semiconductor exports to China to limit its access to advanced computing power.
These new measures underscore the U.S. commitment to securing its technological advantage, enhancing AI security, and building international partnerships based on shared standards and values. The rule aims to balance technological collaboration with safeguarding against national security threats posed by unauthorized or adversarial access to AI systems.
Image-by-Frank-Rietsch-from-Pixabay
