The White House is considering a policy that would require cloud computing companies to report customers²¹ who purchase large amounts of computing power to the government. The policy could help the U.S. identify potential AI threats, especially from foreign entities.

What's going on here?

White House could force cloud companies to disclose AI customers that are using compute resources over a certain limit.

What does this mean?

The policy would force cloud companies like Microsoft, Google, and Amazon to disclose when a customer buys computing resources above a certain level. Though Microsoft, Open AI, and the RAND Corporation think tank have been suggesting variations of this “know your customer” policy, the implementation also risks compelling cloud companies to spy on customers who are competitors.

The case for this policy is providing the US government with early warnings about potential threats from AI systems built by foreign entities.

Why should I care?

This policy matters because computing power is key for developing capable AI systems. Requiring disclosure treats compute like a national resource to be monitored. This proposal could help prevent adversaries from building dangerous AI unchecked. However, if not implemented properly, it risks becoming an AI surveillance program.