Chethana Janith, Jadetimes Staff
C. Janith is a Jadetimes news reporter covering science and geopolitics.
The regulations cap a four-year Biden administration effort to hobble China’s access to advanced chips that can enhance its military capabilities.
On Monday, the U.S. government announced plans to tighten restrictions on the export of artificial intelligence (AI) chips and related technologies. The move seeks to divide global access, ensuring that advanced computing power remains primarily within the U.S. and its closest allies while further restricting China's capabilities.
Under the updated regulations, the number of AI chips that can be exported to most nations will be limited, while U.S. allies will retain unrestricted access to American AI technology. Meanwhile, exports to countries like China, Russia, Iran, and North Korea will remain strictly prohibited.
These comprehensive new rules, revealed as President Joe Biden's administration approaches its final days, extend beyond China and aim to secure the U.S.'s dominant position in AI by tightening global oversight.
"The U.S. is currently leading in AI development and chip design, and it's essential we sustain that leadership," said U.S. Commerce Secretary Gina Raimondo.
The updated measures cap a four-year effort by the Biden administration to curb China's access to cutting-edge chips that could bolster its military and technological capabilities. They also reinforce U.S. leadership in AI by addressing existing loopholes and introducing stricter controls over the global flow of AI technologies.
It remains uncertain how the incoming administration of President-elect Donald Trump will implement the newly announced rules, but both the current and future administrations share concerns about China's growing competitive threat. The regulations are set to take effect 120 days after publication, providing the Trump administration time to evaluate and potentially adjust them.
The new measures will impose restrictions on advanced graphics processing units (GPUs), which are essential for powering the data centers that train AI models. These GPUs are predominantly manufactured by Nvidia, based in Santa Clara, California, with Advanced Micro Devices (AMD) also producing AI chips. Both Nvidia and AMD saw their stock prices drop by 2% to 3% in premarket trading on Monday.
Major cloud service providers, including Microsoft, Google, and Amazon, will have the opportunity to apply for global authorizations to establish data centers. This provision, a significant feature of the new rules, exempts these companies from the country-specific quotas on AI chip exports. Despite this, the shares of all three companies dipped by about 1% during premarket trading.
Authorized companies will need to meet strict conditions, including adhering to stringent security protocols, fulfilling reporting requirements, and demonstrating a commitment to respecting human rights.
Previously, the Biden administration had already enacted broad restrictions on China’s access to advanced chips and the equipment required to produce them. These controls were updated annually to tighten oversight and address countries at risk of diverting technology to China.
Given the significant impact these rules will have on the global AI chip and data center landscape, they have drawn criticism from prominent industry players even before their official release.
Nvidia criticized the regulations as “excessive overreach,” arguing that they restrict “technology already widely available in mainstream gaming PCs and consumer hardware.” Similarly, data center provider Oracle warned earlier this month that the new rules risked conceding “most of the global AI and GPU market to our Chinese competitors.”
The new rules introduce global licensing requirements for certain advanced chips, with some exceptions, while also implementing controls on "model weights" for the most sophisticated "closed-weight" AI models. Model weights, which guide decision-making in machine learning, are considered the most valuable components of AI systems.
The regulations categorize the world into three tiers. Approximately 18 countries, including Japan, the United Kingdom, South Korea, and the Netherlands, will largely be exempt from these restrictions. Around 120 nations, such as Singapore, Israel, Saudi Arabia, and the United Arab Emirates, will be subject to country-specific caps. Meanwhile, arms-embargoed countries like Russia, China, and Iran will be entirely prohibited from accessing the technology.
U.S.-based cloud service providers, including AWS and Microsoft, that are likely to receive global authorizations will face limitations on their deployment of AI computing power. Specifically, they will only be allowed to allocate 50% of their AI capacity outside the United States, no more than 25% in non-Tier 1 countries, and no more than 7% in any single non-Tier 1 nation.
While AI holds immense promise to improve access to healthcare, education, and food, it also presents risks, such as enabling the development of biological and other weapons, supporting cyberattacks, and facilitating surveillance and human rights violations.
“The U.S. must be prepared for the rapid advancements in AI capabilities in the coming years, as they have the potential to profoundly impact both our economy and national security,” said U.S. National Security Adviser Jake Sullivan.
Comentarios