In order to maintain cutting-edge computing power in the United States and among its allies while figuring out further ways to prevent China’s access, the U.S. government announced on Monday that it would further limit exports of artificial intelligence chips and technology.
While retaining a ban on exports to China, Russia, Iran, and North Korea, the new rules would limit the quantity of AI chips that may be shipped to the majority of nations and grant America’s closest friends unrestricted access to American AI technology.
Beyond China, the extensive new regulations announced in the last days of departing President Joe Biden’s administration are intended to help the United States maintain its leadership position in AI by regulating it globally.
According to U.S. Commerce Secretary Gina Raimondo, “it is vital that we keep the U.S. leading AI — both AI development and AI chip design.”
The rules end a four-year Biden administration initiative to restrict China’s access to cutting-edge semiconductors that could improve its military prowess and aim to keep the United States at the forefront of artificial intelligence by plugging holes and erecting new barriers to regulate the flow of chips and the advancement of AI worldwide.
The two administrations have similar opinions about China’s competitive challenge, even though it is uncertain how President-elect Donald Trump’s incoming administration will implement the new regulations. The Trump administration will have 120 days from the date of publication to comment on the rule.
Advanced graphics processing units (GPUs), which power the data centers required to train AI models, will be subject to new restrictions. The majority are produced by Nvidia, a company based in Santa Clara, California, however AI chips are also sold by Advanced Micro Devices.In Monday’s premarket trade, Nvidia and AMD saw declines of 2% to 3%.
Major cloud service providers, such as Microsoft, Google and Amazon, will be able to seek global authorizations to develop data centers, a strong aspect of the new laws that will exempt their projects from the country limitations on AI chips. In premarket trading, the shares of all three firms fell by roughly 1%.
Authorized businesses must adhere to severe guidelines and limitations in order to receive a stamp of approval, such as security standards, reporting obligations, and a human rights policy or history.
Up until now, the Biden administration had placed broad limitations on China’s access to cutting-edge chips and the machinery needed to manufacture them. The controls were updated yearly to strengthen the prohibitions and identify nations that could be transferring the technology to China.
Strong industry voices opposed the idea even before it was released because the regulations would change the global landscape for AI chips and data centers.
The White House would be cracking down on “technology that is already accessible in mainstream gaming PCs and consumer devices,” according to Nvidia, which referred to the rule as “sweeping overreach” on Monday. Oracle, a data center operator, claimed last month that the regulations will give “our Chinese competitors” the majority of the global AI and GPU industry.
With some exceptions, the regulations enforce global licensing requirements for specific chips and establish limits on the so-called “model weights” of the most sophisticated “closed-weight” AI models. Model weights are typically the most valuable components of an AI model and aid in determining decision-making in machine learning.
Three stages are established by the regulation. Japan, Britain, South Korea, and the Netherlands are among the roughly eighteen nations that will essentially be free from the regulations. Singapore, Israel, Saudi Arabia, and the United Arab Emirates are among the 120 additional nations that will be subject to national limitations. Additionally, nations like China, Iran, and Russia who are under an arms embargo will not be allowed to obtain the technology at all.
AWS and Microsoft, two U.S.-based companies that are likely to be granted worldwide authorizations, will only be permitted to use 50% of their total AI computing capacity outside of the US, no more than 25% outside of Tier 1 nations, and no more than 7% in any one non-Tier 1 nation.
In addition to its potential to improve access to food, healthcare, and education, artificial intelligence (AI) can also aid in the development of biological and other weapons, promote cyberattacks, aid in surveillance, and contribute to other violations of human rights.
According to U.S. national security adviser Jake Sullivan, “the U.S. has to be prepared for fast growth in AI’s capability in the coming years, which could have revolutionary influence on the economy and on our national security.”
