The White House today released a document that outlines how the government should use artificial intelligence in national security initiatives.
The new national security memorandum, or NSM, is the fruit of an AI-focused executive order that President Joe Biden signed last year. Alongside the creation of the memorandum, the order launched an array of other machine learning initiatives in the federal government. Some are designed to improve AI safety, while one program will use the technology to find and fix flaws in critical software.
The first focus of today’s NSM is the way the government procures AI technologies for national security missions. Federal agencies will be required to streamline their procurement processes in this area, including by placing a bigger emphasis on buying products that interoperate with one another. When technology products interoperate well out of the box, setting them up requires less time and effort.
Another section of the NSM calls on agencies to prioritize AI when building government supercomputers. Many of the supercomputers that the government has commissioned in recent years, including the exascale Frontier system, include graphics cards to speed up machine learning workloads. The NSM states that AI should likewise be a priority in the development of other emerging technology systems.
Many of the initiatives that the memorandum outlines focus on the private sector. According to the White House, the NSM directs the government to take steps that will improve the security and diversity of chip supply chains. Additionally, the document makes detecting espionage efforts focused on the U.S. AI industry a “top-tier intelligence priority.”
The White House elaborated that the document “directs relevant U.S. government entities to provide AI developers with the timely cybersecurity and counterintelligence information necessary to keep their inventions secure.”
The NSM covers a number of other areas as well. It instructs the government to work with allies on the development of an international framework for ensuring that AI systems are safe, secure and trustworthy. Additionally, the NSM directs the National Economic Council to prepare an assessment about the U.S. AI industry’s competitiveness.
In conjunction with the release of the memorandum, the White House published a framework that provides federal agencies with guidance on how to implement the new requirements. The latter document covers, among other topics, the way officials should go about addressing AI risks.
Lastly, NSM specifies that the government is doubling down on an existing program called the National AI Research Resource. It’s an initiative designed to provide scientists with access to compute infrastructure, datasets and other resources necessary for AI research.