As the summer begins in Washington, D.C., members of U.S. Congress are continuing their cautious, methodical explorations of possible rules to govern artificial intelligence systems. Unlike in the EU, where the AI Act will cover an array of high-risk domains at once, the U.S. approach to AI regulation has been sectoral, siloed and scattershot.
My colleague, IAPP Principal Researcher, Privacy Law and Policy, Müge Fazlioglu, CIPP/E, CIPP/US, previously analyzed the U.S.'s many-pronged approach to AI governance at the federal level. Policymakers have continued to explore these avenues, but one recent change could help to clarify their direction of travel. After all, at least in the Senate, legislators now carry with them a "roadmap," newly minted by their leadership.
On 15 May, the bipartisan Senate AI Working Group released the long-awaited roadmap for AI policy in the U.S. Senate, titled "Driving U.S. Innovation in Artificial Intelligence." The working group consists of Senate Majority Leader Chuck Schumer, D-N.Y., along with his colleagues Sens. Mike Rounds, R-S.D., Martin Heinrich, D-N.M., and Todd Young, R-Ind. IAPP Staff Writer Caitlin Andrews provided initial coverage of the roadmap.
As the title implies, the roadmap is often concerned more with preserving U.S. leadership and innovation around AI systems than with building guardrails around them.
A variety of civil society critics have argued the roadmap lacks real focus on mitigating AI's risks. As Roll Call reported, groups like Color of Change are disappointed by the lack of specificity around addressing how technology might further marginalize already marginalized groups. The American Civil Liberties Union also expressed concerns, stating the roadmap gives little acknowledgment to the risks to civil rights and civil liberties posed by AI, whether by reinforcing existing protections or making new ones focused on AI systems.
The Washington Post reported Sen. Young said the working group embraced vagueness where necessary to come to an agreement.
The policy community was eagerly anticipating the roadmap, not because it would solve AI governance questions but because it could provide a forceful and directional indicator for Senate actions on AI. It has achieved its goal from that perspective, but it takes a careful read to spot the areas of policymaking where Sen. Schumer and his colleagues wish to focus efforts that would reign in AI technologies.
The word "legislation" appears 32 times in the roadmap but there is rarely an explicit call for guardrails. When it does appear, it is often balanced, in keeping with the overarching recommendation for the relevant Senate committees to "develop legislation to leverage public-private partnerships across the federal government to support AI advancements and minimize potential risks from AI."
Nevertheless, committee work on AI is already beginning to pick up steam.
According to The Washington Post, the Senate Rules Committee is considering three election-related AI bills. Although Schumer's roadmap does not explicitly call for legislation in this area, it does encourage committees to "advance effective watermarking and digital content provenance as it relates to AI-generated or AI-augmented election content." Last month, the Senate Committee on the Judiciary Subcommittee on Privacy, Technology, and the Law hosted a hearing on election deepfakes.
Meanwhile, the Senate Committee on Commerce, Science and Transportation, which has authority over rules governing consumer-facing AI systems, is also expected to hold a markup of several bills next week.
What other AI governance efforts does the roadmap express support for, even if only in tentative and tenuous language?
Federal Data Privacy Law
Notably, the Schumer roadmap underscores the urgency for a robust federal data privacy law. It calls upon the Senate committees to craft legislation that ensures the protection of personal information through stringent data minimization standards, a focus area that clearly echoes the approach of the American Privacy Rights Act.
The draft APRA also has multiple AI governance provisions, including prohibitions against algorithmic discrimination, a requirement to conduct design evaluations when designing AI to make consequential decisions and a right to opt out of the use of AI for consequential decisions.
Child online safety
The roadmap also takes the opportunity to express support for ongoing youth safety efforts, highlighting the risks AI-powered systems pose to children. It recommends the development of legislation that establishes clear guidelines and safety protocols for digital platforms to follow, ensuring that young users are shielded from harmful content and predatory behaviors facilitated by AI technologies.
Of course, the Senate Commerce Committee has already advanced the Kids Online Safety Act, which now has more than 60 co-sponsors despite continued vocal opposition. The Children's Online Privacy Protection Act 2.0 was also advanced from committee last summer, though the Senate version differs significantly from the two versions of the framework that share the name in the House.
Financial services
In the financial sector, the roadmap appears to support legislation that would mandate the use of accurate and representative data in AI models. This is to prevent biases and ensure fairness in automated decision-making processes. Senate committees are encouraged to collaborate with industry experts and consumer protection agencies to develop standards that financial services providers must adhere to.
The Senate Committee on Banking, Housing and Urban Affairs hosted a hearing about financial services and AI in September 2023, and another hearing about AI in housing in January.
Content creation and publishing
Addressing the concerns of professional content creators and publishers, the roadmap calls for legislative measures that protect intellectual property rights in the digital age. Committees are tasked with exploring how AI can be used responsibly while respecting the creative outputs and economic interests of individuals and organizations in the content creation industry.
Ban on social scoring
Senator Schumer's roadmap explicitly calls for a ban on the use of AI for social scoring systems. Such legislation would prevent the misuse of AI in evaluating individuals' social behaviors, which could lead to discriminatory practices and breaches of privacy. The Senate committees are expected to consider the ethical implications of AI and enforce a prohibition on social scoring.
AI in health care
The roadmap recognizes the transformative potential of AI in healthcare while stressing the need for appropriate guardrails. It suggests legislation should be introduced to protect patients and provide transparency regarding the use of AI in medical products.
This includes establishing safety measures and ensuring providers and the public are well-informed about the AI systems in operation. In February, the Senate Committee on Finance hosted a hearing on AI and health care.
Transparency requirements for AI systems
A coherent approach to public-facing transparency requirements for AI systems is another legislative priority highlighted in the roadmap. The committees are urged to consider how best to inform the public about the AI systems they interact with, including the logic behind decisions made by these systems. Transparency bills are said to be on the agenda in both the elections context and the consumer protection context.
Content provenance for generative AI
The roadmap incentivizes providers of generative AI software and even hardware products like cameras and microphones to offer content provenance information. Such a legislative step would aim to combat misinformation and ensure the origins of digital content can be reliably traced, fostering a more trustworthy digital environment. Multiple bills and committees have touched on provenance issues and legislation is expected from the Senate Commerce Committee at some point.
Despite the strong interest at the federal level to move forward on AI legislative initiatives, time is running exceedingly short to prepare and pass any major legislation this term.
In the meantime, U.S. state policymakers are not waiting for federal action. The IAPP recently released a state AI governance legislation tracker to help professionals keep up with the shifting expectations.
Just like checking our mirrors while driving a car, it is important for AI governance professionals to stay mindful of roadmaps, guardrails and other emerging obstacles — at the state, federal and global level — even as they keep their eyes on the road.
Please send feedback, updates and new maps to cobun@iapp.org.
