Introduction and Executive Summary

As artificial intelligence (AI) has shown rapid advancements in capabilities over the past year, governments have realized the need for intervention to oversee this technology. In the US, President Biden’s October 30, 2023 Executive Order has been the most sweeping government action to date, stipulating a significant increase in AI activity across Executive Branch agencies. [1]

However, industry leaders and politicians have suggested that existing agencies might not be sufficient to tackle the unique risks AI poses, that the US should instead create a new agency to govern AI. OpenAI CEO Sam Altman made such a suggestion at a Senate Judiciary hearing, and numerous Congresspeople—including Senators Blumenthal and Hawley, and Senators Bennet and Welch, among others—have called for the formation of an AI agency in bipartisan bills. [2]

This paper explores possible paths to an AI agency: the forms in which an agency can exist, the roles an agency can play, and the forces and events that may precipitate an agency’s establishment. After providing a brief overview of US government agencies, this paper details the origin stories of EPA, DHS, CISA, and the FCC Space Bureau (along with nine additional cases briefly outlined in the appendix) to illuminate the intricacies of creating a new agency. 

The case studies suggest that larger agencies tend to be more difficult to establish (regardless of the level of support for the agency among the public and politicians) and can end up creating confusion if given an unclear scope, whereas smaller agencies are easier to create and tend to be understandably specific in their mandates. Agencies also tend not to be created from scratch, instead being formed by combining preexisting government units. Finally, agencies are often formed in response to a clear, identifiable crisis, although many are also preceded by long term social or political movements that sow the seeds for the agency’s formation. 

Policymakers can learn from previous agency origin stories to make an informed decision about whether to create a new agency in the first place, and if so, what type of agency to create, what aspects of AI it should govern, and how the agency should be established. 

Read the full piece here.

Footnotes

1. “FACT SHEET: President Biden Issues Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence,” White House Briefing Room, October 30, 2023, https://www.whitehouse.gov/briefing-room/statements-releases/2023/10/30/fact-sheet-president-biden-issues-executive-order-on-safe-secure-and-trustworthy-artificial-intelligence/.

2. Brian Fung,“Mr. ChatGPT goes to Washington: OpenAI CEO Sam Altman testifies before Congress on AI risks,” CNN, May 16, 2023, https://www.cnn.com/2023/05/16/tech/sam-altman-openai-congress/index.html; “Blumenthal & Hawley Announce Bipartisan Framework on Artificial Intelligence Legislation,” Richard Blumenthal Newsroom, September 8, 2023, https://www.blumenthal.senate.gov/newsroom/press/release/blumenthal-and-hawley-announce-bipartisan-framework-on-artificial-intelligence-legislation;

   “Bennet, Welch Reintroduce Landmark Legislation to Establish Federal Commission to Oversee Digital Platforms,” Michael Bennet Press Releases, May 18, 2023, https://www.bennet.senate.gov/public/index.cfm/2023/5/bennet-welch-reintroduce-landmark-legislation-to-establish-federal-com
   mission-to-oversee-digital-platforms.