The Coming Year of AI Regulation in the States
As we enter 2025, American AI policy is poised to be shaped predominantly by state legislative proposals rather than initiatives from the federal government. With Congress preoccupied with a variety of pressing matters, statehouses are likely to become the epicenter of significant developments in AI regulation.
State Legislative Landscape
In 2024, numerous AI policy proposals were introduced across various states, but only a fraction successfully passed. The majority of these bills were relatively mild, focusing on issues like protecting against malicious deepfakes and forming committees to study AI policy. Notable exceptions included:
- California’s AI Transparency Bill
- Colorado’s Civil-Rights-Based Bill
Expectations for Major Proposals
In the upcoming year, we can anticipate a surge in more robust AI regulatory proposals resembling those seen in the European Union. For instance, legislators in New York are reportedly drafting proposals akin to California’s vetoed SB 1047, which sought to impose liability on AI developers for misuse of their models.
Furthermore, Texas Representative Giovanni Capriglione has introduced the Texas Responsible AI Governance Act (TRAIGA), characterized as a “red state model” for AI regulation, despite its similarities to blue state proposals.
Implications of Proposed Bills
The proposed legislation generally requires AI developers and deployers to conduct detailed algorithmic impact assessments and implement risk management frameworks prior to the release of AI systems. This requirement is particularly relevant for systems used in “consequential decisions” across various industries including:
- Financial Services
- Healthcare
- Insurance
- Business Practices such as Hiring
While such assessments may be useful for narrow AI applications, applying them to generalist models like ChatGPT poses significant challenges. Generalist AI systems have a wide array of potential applications, complicating compliance with the proposed regulations.
Potential Consequences
The broad applicability of these regulations could deter businesses from adopting AI technologies, potentially stifling innovation within the AI sector. For example, contractors in fields as varied as plumbing or electrical work might be required to conduct assessments even for simple tasks like drafting invoices.
State responses to these regulations have varied. For instance, Colorado’s SB 205 was signed into law but with noted reservations regarding its compliance complexity. The intricate nature of these bills could result in a fragmented regulatory landscape, complicating compliance for businesses operating across state lines.
The Risk of a Patchwork Approach
Efforts to uniform AI regulations across states may inadvertently result in a patchwork of laws that vary significantly from one state to another. This could lead to a regulatory environment more complex and ambiguous than the European Union’s AI Act, particularly when combined with the United States’ litigation-heavy culture.
Conclusion
The regulatory landscape for AI in the United States is at a critical juncture. As states prepare to introduce more comprehensive regulations, the federal government may need to step in to provide a cohesive framework that addresses the challenges posed by existing proposals. The outcome of this regulatory evolution will likely shape the future of AI development and deployment across the nation.
