Following Europe, the USA is also adopting stricter rules for companies developing AI technology and for cloud computing providers. From now on the government wants to know what it is doing.

The playing time for OpenAI, Anthropic and Co. is over. Following the example of the European AI Act, Washington will now also introduce new rules for companies that develop AI technology. But while it could take several years in Europe for the rules to actually come into force, in the US it could happen as early as this week, writes Wired. All that is required is a signature from President Biden.

Defense Production Act

Importantly, the law gives the U.S. government more direct control over what AI companies do. The government is relying on the Defense Production Act, a law created in the 1950s after the Korean War that allows the government to monitor private companies if it is in the interests of national security. In this context, that means AI companies must inform the government of any project that requires a “significant amount of computing power.”

This would give the government access to sensitive data and programs from American AI companies. The extension of the Defense Production Act to the AI ​​industry is therefore special. AI may well have military applications, even if the major players try to avoid using their technologies for this purpose as much as possible. They also have to undergo regular security tests.

Know your customers

The proposed laws affect cloud providers that host AI models. They need to be more vigilant about who they allow access to their cloud infrastructure, especially when it comes to foreign actors. The government applies a “Know Your Customers” regulation here, requiring providers to thoroughly verify the identity of non-American customers who register. There will be minimum requirements for identifying foreign users and cloud computing companies will also have to prove annually that they respect these.

In principle, the rules apply to all foreign users of American cloud computing services, but more specifically the legislation is intended to prevent Chinese companies from entering into a cloud contract. “We cannot allow non-state actors or China or people we don’t want to have access to our cloud to train their models,” Economy Secretary Gina Raimondo told Reuters. The law is part of a series of measures to hold China back in the AI ​​race.