US tightens AI export controls to curb hostile exploitation

To solidify the United States’ position as a world leader in artificial intelligence (AI), the Biden-Harris administration has announced an interim final rule on AI adoption. The policy hopes to improve national security and economic strength while ensuring the responsible global deployment of U.S. AI technology.

“In the wrong hands, powerful AI systems could be used to develop weapons of mass destruction, support powerful cyber operations, support human rights abuses, and more,” the White House said in a press statement. potentially exacerbating significant national security risks, such as mass surveillance.Currently, countries of concern are actively adopting AI (including U.S.-made AI), and U.S. AI leadership. trying to undermine it.”

The interim final rule builds on previous export controls introduced in October 2022 and October 2023. This rule was developed through extensive engagement with industry leaders, bipartisan lawmakers, and international allies.

Main provisions of AI dissemination rules

This rule streamlines licensing for chip exports, raises AI security standards, and establishes a mechanism to responsibly distribute U.S. technology around the world. It will also introduce guardrails to limit access by countries of concern.

The six core mechanisms that underpin this policy are:

Unlimited Sales to Key Allies: Chip sales to 18 allies will remain unrestricted, allowing countries with strong security and technology consistent with U.S. national interests to seamlessly purchase. Streamlined, low-risk orders: Orders containing up to 1,700 advanced GPUs commonly used in universities, medical institutions, and research institutions can avoid licensing requirements and ship for non-sensitive purposes. Speed ​​it up. Universal Verified End User (UVEU) Status: Entities in allied countries that meet rigorous security standards can earn UVEU status, giving them access to up to 7% of the world’s AI computing power in other countries while maintaining cutting-edge development within the United States. can be placed. User (NVEU) status: Trusted entities outside the UVEU framework can purchase compute power equivalent to 320,000 advanced GPUs over two years to support local and regional AI initiatives. Violate security. Non-VEU purchases: Organizations outside of our close alliance network can purchase up to 50,000 advanced GPUs per country, ensuring access to U.S. technology for critical services like healthcare and communications. Intergovernmental Agreement: Countries that align their technology security and AI governance efforts to U.S. standards can double chip purchase limits to 100,000 GPUs and foster an international ecosystem of shared values. Masu.

Target exploitation by adversaries

The rules tighten restrictions on countries of concern, aimed at thwarting their efforts to exploit advanced AI systems. The main measures include:

Ensure that advanced semiconductors sold overseas are not used to train hostile AI systems. Restrict model weight transfer for closed-weight AI models to untrusted actors while protecting open-weight model research. Set strict security standards for storing and using advanced AI models to prevent unauthorized access.

all or nothing

“One of the most frustrating things about legislation from any administration is that it tends to be all-or-nothing. We need regulation, but we do consider access, surveillance, and the use of AI. ” says Kris Bondi, CEO and co-founder of Mimoto.

“While I agree that the use and protection of AI is important to U.S. national security and economic strength, this form of isolationism undermines innovation. All progress is made on American soil. The bubble it creates will limit, rather than protect, our ability to evolve and compete on a global scale,” she added.

maintain U.S. and Allied supremacy

Bugcrowd founder Casey Ellis said the rules reflect a broad consensus in Washington that AI is establishing itself as a “powerhouse” technology. Maintaining the advantage of the United States and its allies in this area is key to maintaining America’s status as a global superpower. The regulation’s emphasis on not offshoring this critical technology highlights the strategic importance of AI in shaping future economic and geopolitical power dynamics.

“Historically, America’s advantages in AI and semiconductor technology have been driven by its ability to rapidly innovate and compete globally. Overly restrictive export controls have alienated allies and “There is a risk that the United States could be prevented from accessing important markets, potentially weakening U.S. technological advantages,” Ellis added.

“However, the need for strategic constraints remains clear, especially to prevent adversaries like China and Russia from weaponizing advanced AI capabilities against the United States and its allies. The goal is to precisely target these limits to high-risk technologies without compromising economic opportunity or innovation.”

AI capabilities serve civil and military purposes and have immediate national security implications that justify government surveillance. “The current approach is very similar to the export control regulations imposed on cryptographic technologies, making it a critical balancing act to protect national interests while enabling innovation,” Ellis explained.

Balance sharing and security

Stephen Kowski, field CTO at SlashNext, commented that the rules hope to strike a balance between protecting advanced AI capabilities and maintaining technological leadership. “Given the sophistication of cyber threats and the potential for misuse of AI systems, it is important to ensure the security of AI infrastructure and computing resources. prevent it from being used in ways that compromise security or enable malicious activity.”

Kowsky said technology sharing must be balanced with robust security controls and verification systems to prevent misuse. Smart partnerships with trusted allies can foster innovation while maintaining critical safeguards against threats. “The key is to put in place precise and targeted controls rather than broad restrictions.”