AI remains a critical focus of both the federal government and industry, with multiple efforts in recent weeks to address governance of the development and use of AI in the United States. On February 26, 2024, a U.S. Department of State-commissioned report, titled "Defense in Depth: An Action Plan to Increase the Safety and Security of Advanced AI" (Action Plan), proposed multiple U.S. Government and partner nation lines of effort to address growing national security risks posed by rapidly expanding AI capabilities, including the expectation of achieving artificial general intelligence (AGI). Just days later, on March 5, 2024, House Chairman Comer and Ranking Member Raskin introduced the Federal AI Governance and Transparency Act. This bipartisan bill would focus government resources on increasing transparency, oversight, and responsible use of federal AI systems and centrally codifying federal governance of agency AI systems. Additionally, on March 28, 2024 the Office of Management and Budget released the final guidance on Memorandum M-24-10,Advancing Governance, Innovation, and Risk Management for Agency Use of AIas we address here.

The State Department-commissioned Action Plan broadly warns that the federal government must move "quickly and decisively" to avert substantial national security risks stemming from AI. The Action Plan outlines five lines of effort (LOEs) designed to put the government on a path to stabilize (LOE1), strengthen (LOE2, LOE3), and scale (LOE4, LOE5) advanced AI development safely and securely.

  1. LOE1, titled "Establish Interim Safeguards to Stabilize Advanced AI Development," focuses on potential executive branch actions to lower AI risk in the near term (1-3 years), while setting the conditions for successful long-term AI safeguards. Examples in the Action Plan include creating an AI Observatory, mandating an interim set of responsible AI development and adoption (RADA), and creating an interagency AI Safety Task Force (ASTF) to coordinate implementation and oversight of RADA safeguards.
  2. LOE2, titled "Strengthen Capability and Capacity for Advanced AI," outlines specific actions that the federal government could take to increase its preparedness for rapidly addressing incidents related to advanced AI development and deployment. For example, the Action Plan recommends coordinating the development of an indications and warnings (I&W) framework for advanced AI and AGI incidents.
  3. LOE3, titled "Increase National Investment in Technical AI Safety Research and Standards Development," offers recommendations that the federal government could take to strengthen domestic technical capacity in advanced AI safety and security, AGI alignment, and other technical AI safeguards. These actions include directly funding advanced AI safety and security research and promulgating safety and security standards for responsible AI development and adoption.
  4. LOE4, titled "Formalize Safeguards for Responsible AI Development and Adoption by Establishing an AI Regulatory Agency and Legal Liability Framework," focuses on specific actions the legislative branch could take to establish the conditions for long-term (4+ years) domestic AI safety and security such as creating the Frontier AI Systems Administration (FAISA), a regulatory agency that would have rulemaking and licensing powers to oversee AI development and deployment.
  5. Lastly, LOE5, titled "Enshrine AI Safeguards in International Law and Secure the AI Supply Chain," suggests near-term diplomatic actions and longer-term measures the federal government could take to establish an effective AI safeguards regime in international law while securing the AI supply chain. Recommendations from the Action Plan include building a domestic and international consensus on catastrophic AI risks and safeguards and establishing an International AI Agency (IAIA) to monitor and verify adherence to those safeguards.

The Action Plan also recommends establishing civil and criminal liability for "dangerous behaviors" by individuals and entities involved in the AI supply chain. For example, the report suggests that failing to accurately report high-performance AI hardware to the FAISA or responding to information requests from the FAISA with misleading data may constitute misdemeanors, and disregarding an emergency order to halt AI development activities or breaching the conditions of a license may constitute a felony.

Similarly, the proposed Federal AI Governance and Transparency Act focuses on creating federal standards, consolidating other existing laws that impact AI, and establishing AI transparency and accountability. Specifically, the bill focuses on the following key objectives:

  1. defining federal standards for responsible AI use by codifying, in law, key safeguards for the development, acquisition, use, management and oversight of AI used by federal agencies;
  2. strengthening governmentwide federal AI use policy authority and requirements, by re-codifying and clarifying the role of the Office of Management and Budget in issuing governmentwide policy guidance, in concert with existing federal IT and data policy requirements;
  3. establishing agency AI governance charters, which would require the publication of governance charters for high-risk AI systems and other AI systems used by federal agencies that interact with sensitive personal records covered by the Privacy Act;
  4. creating additional public accountability mechanisms by establishing a notification process for any individual or entity that has been substantively and meaningfully affected by an agency determination influenced by AI;
  5. streamlining and consolidating existing law regarding the government's use of AI and repealing repetitive provisions in the AI in Government Act of 2020 and the 2022 Advancing American AI Act; and
  6. updating existing Privacy Act Personally Identifiable Information (PII) record notice requirements and FAR procurement rules.

The House Committee on Oversight and Accountability has considered this bill and voted on a markup. Both the Action Plan and the Federal AI Governance and Transparency Act appear to be in furtherance of Executive Order 14110, Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.

Government contractors engaged in AI development or use should be aware of the Action Plan and recent bill report because, taken together, both suggest broad changes to the federal government's and industry's current approach to AI. Additionally, the bill in particular aims to increase U.S. participation from stakeholders, such as government contractors, in critical AI-based activities.

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.