AI Chip Gatekeeping: The Role of Major Cloud Providers Outside the US

In a notable move against Beijing, the United States plans to enable companies like Google and Microsoft to serve as gatekeepers for global access to valuable AI chips, according to two sources aware of the draft plan.

This proposed scheme is set to be revealed soon and includes strict requirements for these companies. They would need to report essential information to the US government and prevent access to AI chips for Chinese entities.

With this arrangement, these companies could provide artificial intelligence capabilities via the cloud to international clients without requiring a license, as explained by the sources.

This new initiative demonstrates the urgency felt by officials in the last days of the Biden administration. They are trying to simplify the process for approving the export of AI chips while also ensuring that they do not fall into the hands of malicious actors.

The United States is particularly wary of China potentially leveraging AI technologies to enhance its military power, orchestrate cyberattacks, or even develop bioweapons.

The US Commerce Department has refrained from commenting on the specific details and timeline of the proposed regulations, and sources have warned that these plans may evolve.

Google and Microsoft have not yet responded to inquiries for comments.

This measure is reminiscent of a national security agreement made by Microsoft in April with the US government, which allowed them to provide AI technology to the Emirati company G42.

According to the new draft regulations, other companies without gatekeeper status will also have the opportunity to apply for licenses to import a limited number of high-end AI chips from Nvidia and AMD in various countries.

Nvidia, known for producing some of the most powerful AI chips worldwide, has expressed readiness to collaborate with the administration regarding these new rules, while AMD has not provided an immediate response to requests for comments.

Seventeen allied nations, such as the Netherlands, Japan, and Taiwan are set to be exempt from these caps, allowing them unlimited access to AI chips or the technologies they enable. At the same time, countries that are under nuclear embargo, including Russia, China, Iran, and Venezuela, will remain restricted from acquiring US-made AI semiconductors.

Some nations may express concerns regarding the caps. Geoffrey Gertz, a former White House official currently associated with the Center for a New American Security, noted that a global cap system could result in significant worries from US allies and partners. These nations are cautious about the unilateral stance taken by the US in determining who can access crucial AI chips.

The US government is currently completing a final review of a new rule concerning "Artificial Intelligence Diffusion" as indicated by a recent government announcement, signaling that publication may be imminent. Sources have indicated that this posting relates to the aforementioned AI caps.

Concerns have also been raised by the Information Technology Industry Council, which includes major entities like AMD and Google among its members. They fear that the Biden administration may be hastily pushing this complex rule through without industry input, risking unintended consequences.

Naomi Wilson, the council's senior vice president of Asia and global trade policy, commented, "If the reports are correct, such changes would expand the scope of export controls dramatically and could have substantial global repercussions."

The new regulations build on a program launched in September that allows pre-approved data centers overseas to receive AI chips without needing a license. To achieve this status, data centers must provide detailed information regarding their customers, business operations, access restrictions, and cybersecurity measures.

AI, Chips, Regulations