On May 15th, after nearly a year of study, the Bipartisan U.S. Senate AI Working Group released its Roadmap for Artificial Intelligence (AI) Policy in the United States Senate. Based on insights collected during three educational Senate AI briefings, several committee hearings, and nine AI Insight Forums held in 2023, the Working Group identified several areas of consensus that it believes warrant bipartisan consideration of AI policy and legislation.
The Senate AI Roadmap maintains that existing laws, including those related to consumer protection and civil rights, must consistently and effectively apply to AI systems. It also states that developers, deployers, and users of AI systems must be held accountable under these laws. Additionally, the Roadmap highlights the need for transparency and explainability in AI systems to enable regulators to enforce existing laws effectively.
Additionally, the Roadmap encourages relevant committees to agree on standardized definitions for all key AI terms, to identify any gaps in the application of existing laws to AI systems, and to develop legislation to address such gaps as needed. Notably, the Roadmap includes recommendations for new legislation, including support of a comprehensive federal privacy law.
We’ve summarized below key takeaways from the areas of consensus and policy proposals as outlined in the Senate AI Roadmap.
Supporting U.S. Innovation in AI
The Senate AI Roadmap advocates for continued federal investments in AI, recommending that the executive branch and Senate Appropriations Committee aim for the $32 billion annual spending level proposed by the National Security Commission on Artificial Intelligence (NSCAI). Key priorities include funding cross-government AI research and development (R&D), supporting semiconductor R&D, authorizing the National AI Research Resource (NAIRR), and launching the “AI Grand Challenge” program. The Roadmap also emphasizes modernizing federal internal governance, enhancing cybersecurity, and leveraging public–private partnerships to advance AI innovation and mitigate risks. Additionally, the Roadmap supports measures to help small businesses and startups compete in the AI marketplace and calls for increased access to AI testing tools.
AI and the Workforce
The Senate AI Roadmap acknowledges widespread concerns about AI’s impact on jobs and emphasizes the need to address potential workforce displacement. To address these concerns, the Roadmap advocates for consulting a range of stakeholders during AI development and deployment and for legislation focused on training and upskilling workers. This includes incentives for businesses to integrate new technologies and reskill employees to ensure they can thrive in an “AI-enabled economy.” The Roadmap also calls for exploring policies to mitigate the disruptive effects of AI on employment, including implementing a policy framework and private sector best practices and strategies to combat workforce displacement. The Roadmap supports the Workforce Data for Analyzing and Tracking Automation Act (S. 2138), which would authorize the Bureau of Labor Statistics (BLS) to monitor automation’s impact on the workforce and measure those trends over time.
High-Impact Uses of AI
The Senate AI Roadmap emphasizes the need for existing laws, such as consumer protection, civil rights, and antidiscrimination laws, to apply to AI systems and their developers, deployers, and users, to empower regulators to enforce these laws. The Roadmap encourages committees to develop legislation that ensures regulators can access necessary information about AI systems to determine legal compliance and place appropriate transparency, explainability, testing, and evaluation requirements on high-risk uses of AI on a case-by-case basis. The Roadmap recommends that relevant committees review whether other potential uses of AI should be either limited or banned. The Roadmap also suggests that relevant committees consider developing legislation to establish public-facing transparency requirements for AI systems that align with any potential risk regime and do not inhibit innovation.
Elections and Democracy
The Senate AI Roadmap urges committees and AI developers to promote watermarking and digital content provenance for AI-generated election content. It advocates for robust protections to prevent false AI-generated content while upholding First Amendment rights. The Roadmap recognizes the U.S. Election Assistance Commission (EAC) for its AI Toolkit for Election Officials and the Cybersecurity and Infrastructure Security Agency (CISA) for its cybersecurity resources and encourages states to use these tools to protect elections.
Privacy and Liability
The Senate AI Roadmap emphasizes the issue of accountability for the development and deployment of AI products and the need for protections for personal information used in AI products. It encourages relevant committees to consider clarifying standards and enforcement of such standards to hold developers, deployers, and end users accountable for harm caused by AI technologies. The Roadmap also recommends exploring policies to reduce personal information stored in AI systems and supports a “strong comprehensive federal data privacy law” to protect such personal information.
Transparency, Explainability, Intellectual Property, and Copyright
The Senate AI Roadmap supports transparency surrounding AI systems as they continue to be developed. To further this goal, it recommends considering legislation to govern the disclosure of the use of AI products and information contained in training datasets. It also encourages relevant committees to consider incentives for companies to provide content provenance information, either voluntarily or through legislation. As for intellectual property, the Roadmap advises relevant committees to consider name, likeness, and voice protections as synthetic content becomes more widely available and encourages a review of intellectual property office reports on AI-related works. Lastly, the Roadmap supports efforts toward establishing AI public awareness and educational campaigns.
Safeguarding Against AI Risks
The Senate AI Roadmap recognizes the need to understand and mitigate the risks of AI systems, urging testing, evaluation, and a “resilient risk regime” that balances risk and data protection with the need for AI innovation. The Roadmap encourages relevant committees to support such risk testing and evaluation, including determining when evaluation should be required before the deployment of AI models and developing legislation to advance risk-aware R&D efforts. The Roadmap supports exploring an AI-focused Information Sharing and Analysis Center to aid risk monitoring by promoting communication between private companies and the federal government.
National Security
The Senate AI Roadmap provides several recommendations regarding the national security effects of AI. The Roadmap supports expanding AI workforce capabilities in the armed services, including AI digital engineering training programs, enhanced security clearance efficiency, and improved placement opportunities. The Roadmap recognizes AI use in scientific discovery and bioweapons, emphasizing the need to consider recommendations from relevant organizations to defend against potential threats. More generally, it encourages the development of frameworks for export controls, national security issues surrounding the use of AI systems by adversaries, and management of space debris and the energy demand of AI systems. Lastly, the Roadmap emphasizes the importance of international coordination, including agreements with allies, international AI research institutes, and the cross-border flow of information and digital goods.
Stemming from a “shared recognition of the profound changes artificial intelligence (AI) could bring to our world,” the Senate AI Roadmap provides high-level recommendations for continued efforts by Senate committees to understand AI technology and its effects. We will continue to monitor new updates and publications from the relevant committees and working groups.
Danielle Santos, an AI Analyst, contributed to the drafting of this blog post.