The Biden Administration has issued the first ever National Security Memorandum (NSM) on AI, a technology that it described as “era-defining”.

The administration believes that US security is in danger of being undermined by the growth of AI, and states that the US must maintain its lead in its development. The NSM outlines a number of actions for national security agencies to undertake to prepare the US government for the growing prominence of AI.

Protecting AI-related intellectual property

One measure outlined in the memo is a declaration of US government policy to protect AI-related intellectual property and infrastructure belonging to US industry, civil society, and academia from foreign intelligence threats.

The memorandum especially focuses on the threat posed by the development of artificial general intelligence (AGI), with private sector companies in the US competing against “national champions” in the People’s Republic of China (PRC).

By establishing a “security perimeter” and “defensive umbrella” around the US AI ecosystem, the memorandum said that the US government can help ensure that breakthroughs by American companies in AGI are not stolen by the PRC.

Managing growing energy demands of AI

The memo addresses the growing energy demands of AI development. It directs the White House Chief of Staff to coordinate efforts to streamline permitting, approvals, and incentives for the construction of AI-enabling infrastructure.

The NSM outlines actions to address the need for skilled individuals in AI and related fields. It designates as a national security priority the ability of skilled foreigners to enter and work in the US. It directs an update to government-wide procedures for attracting, hiring, developing, and retaining AI and AI-enabling talent, including attracting talent from industry, academia, and civil society.

The AI Safety Institute

The memorandum establishes the AI Safety Institute (AISI) within the Department of Commerce as a “one-stop shop” where private sector developers can voluntarily submit their frontier AI models, before and after public deployment, to be tested for safety, security, and trustworthiness. The memo also directs the AISI to issue guidance within six months to AI developers on how to test, evaluate, and manage safety risks arising from dual-use foundational models.

In an attempt to speed up the adoption of AI by defence and intelligence communities, the memo addresses the challenge of slow procurement and accreditation of AI capabilities. The Department of Defense (DoD) and the Office of Director of National Intelligence (ODNI) are directed to identify and address issues involved in the procurement of AI by DoD and intelligence entities. Each agency using AI is directed to take steps to accelerate the approval and accreditation of AI systems.

Working with Allies

Within the memo, the administration also issues a directive for departments and agencies, particularly the DoD, to enable co-development and co-deployment of AI capabilities with select allies and partners. It singles out the UK and Australia as particular partners for development of AI for defence and security purposes.