[co-author: Raj Gambhir]
Welcome to this week's issue of AI: The Washington Report, a joint undertaking of Mintz and its government affairs affiliate, ML Strategies.
The accelerating advances in artificial intelligence (“AI”) and the practical, legal, and policy issues AI creates have exponentially increased the federal government’s interest in AI and its implications. In these weekly reports, we hope to keep our clients and friends abreast of that Washington-focused set of potential legislative, executive, and regulatory activities.
This issue covers the inaugural AI Insight Forum held on September 13, 2023, in the United States Senate. We also cover three congressional hearings on AI that took place during the week of the inaugural AI Insight Forum. Our key takeaways are:
- During the AI Insight Forum, assembled stakeholders discussed a variety of AI-related topics, ranging from the dangers of open-sourcing large-language models to the potential impact of AI on elections. This meeting was the first of nine scheduled forums.
- Leader Chuck Schumer (D-NY) hopes that these forums will build “a foundation for bipartisan AI policy that Congress can pass.” However, Schumer has himself acknowledged the continuing importance of congressional hearings to the process of creating AI legislation.
- During key AI hearings held during the same week as the inaugural AI Insight Forum, Senators Richard Blumenthal (D-CT) and Josh Hawley (R-MO) explained their recently announced framework for AI legislation, the Senate Commerce Committee examined a range of AI-related consumer protection bills, and the House Oversight Committee discussed the federal bureaucracy’s adoption of AI tools.
The Inaugural AI Insight Forum
On September 13, 2023, a bipartisan group of senators headed by Senate Majority Leader Schumer held the first of nine scheduled AI Insight Forums, closed-door and Senate-wide meetings with technologists, academics, and other stakeholders on AI.
In his published opening remarks to the forum, Leader Schumer asserted that the purpose of the forum is “building a foundation for bipartisan AI policy that Congress can pass.” He called on Congress to draft legislation that allows for the emergence of “transformative innovation” driven by AI, including “systems to unlock new cures, improve education, protect national security, protect the global food supply.”
According to comments Leader Schumer made to the press following the conclusion of the event, the topics covered during the inaugural AI Insight Forum included:
- The Pros and Cons of Open-Source AI Models: Many software developers and corporations have lauded open-source architecture as a boon to technological progress and economic growth. In the realm of AI, however, some lawmakers and experts fear that allowing large-language models to be open-source presents an intolerable risk. Under an open-source architecture, relatively unsophisticated actors could exploit powerful AI models. In June 2023, Senators Blumenthal and Hawley sent a letter to Meta criticizing their decision to make their LLaMA model open source to select researchers.
- AI’s Potential to Revolutionize Health Care: From publicly available information, it is not clear which potential impacts of AI on health care the members specifically discussed at this forum. Based on Senator Schumer’s comment regarding the ability of AI to help researchers “unlock new cures,” it is likely that the assembled experts and lawmakers discussed the potential for AI to catalyze drug discovery, among other topics.
- AI and Elections: With the 2024 election approaching, many lawmakers have expressed concern over the potential for AI-generated “deepfakes” (videos where a person’s face or body is digitally altered to appear to be someone else, typically used maliciously or to spread false information) to potentially sway election results. This concern has prompted lawmakers in both parties and both chambers of Congress to introduce bills on AI and election security, with proposals ranging from the enforcement of a content labeling requirement on AI-generated content in election advertisements to the outright ban on deceptive AI election–related deepfakes.
- AI-Driven Workforce Disruptions: Many fear that AI-driven automation may eliminate millions of currently existing jobs over the next few decades. To prepare for these transformations, lawmakers have introduced a range of bills, including those that would lay the groundwork for large-scale workforce training and others that focus on building AI expertise within the federal bureaucracy.
- Choice of AI Regulator: Given the potentially transformative impact of AI over the coming decades, some lawmakers, such as Senators Blumenthal and Hawley, have called for the establishment of an “independent oversight body” tasked with carrying out comprehensive AI regulation. Others, such as FTC Chair Lina Khan, have argued that existing agencies possess the competency and experience to regulate AI.
Though Leader Schumer lauded the AI Insight Forum as “one of the most informative discussions ever held in Congress,” he has acknowledged that “committees must continue to be the key drivers of Congress’s AI policy response…” The week of the inaugural AI Insight Forum also saw multiple hearings on AI.
Senate Judiciary Committee Hearing: “Legislating on Artificial Intelligence”
On September 12, 2023, the Senate Judiciary Committee’s Subcommittee on Privacy, Technology, and the Law held a hearing entitled “Oversight of A.I.: Legislating on Artificial Intelligence.” The hearing primarily focused on the framework for an AI bill recently introduced by Chair Blumenthal and Ranking Member Hawley.
The Blumenthal-Hawley AI Framework
Chair Blumenthal and Ranking Member Hawley released an outline for a comprehensive AI bill on September 8, 2023. During the hearing, Chair Blumenthal announced that he and Ranking Member Hawley hope to turn this framework into legislation “at least by the end of this year.”
The Blumenthal-Hawley proposal would establish a licensing regime whereby companies developing “sophisticated general-purpose A.I. models…or models used in high-risk situations” would be “required to register with an independent oversight body.” Senator Blumenthal explained that while existing agencies should play a role in overseeing relevant AI use cases, “there needs to be an oversight entity…so that those harms [caused by AI] can be prevented.”
The framework would grant individuals a private right of action against companies whose AI models “breach privacy, violate civil rights, or otherwise cause cognizable harms.” Current protections limiting platform liability would not apply to AI. The framework would also direct Congress to address the issues of AI-generated deepfakes, child sexual abuse material, and election interference.
To prevent countries engaged in “gross human rights violations” from acquiring advanced AI tools, the framework directs Congress to utilize “export controls, sanctions, and other legal restrictions.” Senator Blumenthal analogized the framework’s approach to high-risk AI models to “nuclear proliferation. We cooperate over safety, in some respects, with other countries, some of them adversaries. But we still do everything in our power to prevent American companies from helping China or Russia in their nuclear programs.”
Companies developing and deploying AI systems would be subject to a suite of transparency requirements, including the labeling of AI-generated content and the watermarking of deepfakes. Finally, the framework asserts that “strict limits should be imposed on generative A.I. involving kids.”
Senate Commerce Hearing: “The Need for Transparency in Artificial Intelligence”
On September 12, 2023, the Senate Commerce Committee’s Subcommittee on Consumer Protection, Product Safety and Data Security held a hearing on “The Need for Transparency in Artificial Intelligence.” During the hearings, senators discussed a range of AI-related consumer protection bills introduced in the 118th Congress.
A Panoply of AI-Related Consumer Protection Proposals
As the impact of AI sweeps broadly, the committee members approached the issue of AI’s potential consumer protection harms from a variety of angles.
Senator Maria Cantwell (D-WA) discussed the need for Congress to invest in workforce training to avoid widespread job loss, asserting that “we need something as grand as the GI Bill was after World War Two in empowering Americans for new opportunities in this area.” In a similar vein, Senator Todd Young (R-IN) raised the issue of workforce modernization within the federal workforce. Along with workforce training, members advocated for measures that would regulate certain classes of content produced by generative AI. Senator Amy Klobuchar (D-MN) discussed her recently introduced bill, the Protect Elections from Deceptive AI Act, which would ban most instances of “fraudulent AI generated content in political ads.” As a means to “respond to [AI generated] deep-fakes,” Senator Cantwell raised the possibility of the federal government establishing a content watermarking regime. Finally, Senator Ben Ray Luján (D-NM) announced that his office is working on a bill that would establish disclosure requirements that would inform consumers “how their information is going to be used [by an AI model] and know when content is generated by AI.”[1]
The Necessity of Enacting Comprehensive Privacy Legislation
Though this hearing principally concerned AI regulation, much discussion during this hearing concerned a different long-standing priority: comprehensive data privacy regulation.
Many commentators have noted parallels between Congress’s current drive towards comprehensive AI regulation and its ongoing and unsuccessful effort to secure a comprehensive data privacy law. Senator Jerry Moran (R-KS) vented his frustration on the issue, expressing that it is “annoying that we’re here now on AI when we’ve been unsuccessful in reaching conclusions on data privacy legislation.”
Primarily, however, members invoked the issue of data privacy regulation to argue that implementing strong federal data privacy protections is an important precursor to, if not a precondition of, federal AI regulation. Chair John Hickenlooper (D-CO), for example, argued that since “AI trains on publicly available data…Congress needs to pass comprehensive data privacy protections” in order to fully address the consumer protection issues attendant with the development of AI.
Developing AI Regulation with an Eye to International Issues
Though members of the committee largely supported the enactment of policies and procedures that would grant consumers protections against AI-related harms, certain members emphasized the need to balance the imperative of consumer protection with that of international competitiveness.
Along with considering competition with adversaries in the field of AI, Ranking Member Marsha Blackburn (R-TN) raised the issue of coordinating AI regulation with allies. Ranking Member Blackburn discussed the frustration felt by some in Europe given the slow development of AI regulation in the United States relative to the European Union’s (“EU”) progress in adopting its Artificial Intelligence Act. The Ranking Member acknowledged that there may be a gap between the full implementation of the EU’s AI Act and the signing into law of a comparable standard in the United States and wondered how businesses “would navigate a patchwork system around the globe when it comes to” AI regulation.
House Oversight Committee Hearing: “How Are Federal Agencies Harnessing Artificial Intelligence?”
On September 14, 2023, the House Committee on Oversight and Accountability's Subcommittee on Cybersecurity, Information Technology, and Government Innovation held a hearing entitled “How are Federal Agencies Harnessing Artificial Intelligence?” During the hearing, committee members discussed strategies for managing the federal bureaucracy's adoption of AI tools. No specific new legislation was the focus of the hearing.
The Peril and Promise of Federal Agencies’ Adoption of AI
Members of the committee expressed optimism regarding the potential of AI to enhance the functioning of federal bureaucracies.
The optimism expressed by committee leadership on this matter was tempered by a recognition of the potential government abuses that AI could facilitate. Ranking Member Connolly argued that the House Oversight Committee “must hold federal agencies accountable to ensure that they’re making appropriate choices about whether and when AI is right for their mission.”
Sufficiency of the Biden Administration’s Executive Branch Actions on AI
While Democratic committee members lauded President Biden’s various AI initiatives, the Republican majority questioned whether the White House’s efforts have been sufficient to facilitate federal agencies’ adoption of AI tools.
Chair Mace repeatedly criticized the Office of Management and Budget (“OMB”) for not yet carrying out its assessment of “federal AI workforce needs” as mandated by the National AI Initiative Act of 2020. The “OMB is now more than two years behind schedule” with regard to this responsibility, noted Chair Mace. The Chair also spoke disapprovingly of the Office of Science and Technology Policy’s “government-wide AI use-case inventory.”
Training an AI-Age Federal Workforce
While committee leadership expressed differing views regarding the Biden administration’s AI initiatives, both Chair Mace and Ranking Member Connolly urged Congress to invest in workforce training measures that would prepare the federal bureaucracy for the demands of the AI age. Ranking Member Connolly commended Chair Mace for her work in getting the AI Training Expansion Act of 2023 voted out of committee. This bill would expand the categories of federal employees that are required to receive training on AI.
[1] During the hearing, Senator Luján explained that this yet-to-be-introduced AI transparency bill would be similar to his proposal on nutrition labeling for broadband.
[View source.]
