The approach to regulating AI in the US remains fragmented. Similar to data protection regulations, individual states are beginning to regulate AI. Despite this fragmented approach and dissimilarities, there are significant similar features to the state-level regulations.
This post provides a glacial view on state of AI legislations in US including the Executive Order, bills and regulations in Utah and California who are among the first states in the US to take steps to regulate AI and AI Technology, as well other states who are making progress towards AI regulations.
US Executive Order on Safe, Secure and Trustworthy AI
The US Executive Order on Artificial Intelligence (AI) was issued by President Joe Biden October 2023.
Principles and Priorities:
The order emphasizes that AI should be developed and used in a manner consistent with American values, including fairness, transparency, and accountability.
It recognizes the importance of maintaining technological leadership while ensuring public trust.
Federal Agency Responsibilities:
Federal agencies are directed to enhance AI research, development, and deployment.
They must prioritize AI applications that address national priorities, such as healthcare, climate change, and national security.
Data and Privacy:
Agencies are instructed to improve access to high-quality and diverse data for AI research and development.
Privacy protections and civil liberties must be considered in AI deployment.
International Cooperation:
The order encourages international collaboration on AI research and standards.
It emphasizes cooperation with allies to address global challenges.
Accountability and Transparency:
Agencies must provide public information about their AI activities.
They should consider the potential impact of AI on employment and the economy.
Ethical AI:
The order calls for the development of AI standards that align with ethical considerations.
It emphasizes avoiding biases and discriminatory practices.
Utah and California on AI - Key Points
Both Utah and California recognize the importance of addressing AI’s impact while fostering innovation. Other states are making progress towards regulating AI.
Utah’s Approach to AI Regulation
AI Policy Act (Senate Bill 149): Utah recently approved the AI Policy Act, becoming one of the first U.S. states to address private-sector AI deployments beyond public-sector requirements. Notable features include:
Consumer Protection Focus: Rather than regulating the technology itself, the act holds businesses accountable if their use of generative AI harms consumers.
Transparency Obligations: Certain licensed professionals (e.g., mental health providers) must disclose interactions with AI technology or materials generated by generative AI.
Penalties: Violations can result in administrative fines (up to $2,500) and civil penalties (up to $5,000). Criminal charges may apply for using generative AI models for criminal activities.
Office of AI Policy: Utah establishes an AI Learning Lab within its consumer protection division. The lab studies AI risks and benefits, makes recommendations, and invites stakeholder input on new technology applications across industries.
California’s Approach to AI Regulation
Executive Order by Governor Newsom: California recognizes its role as a global hub for generative artificial intelligence (GenAI). Governor Newsom signed an executive order to study AI development, use, and risks throughout the state. Key points include:
Measured Approach: California aims to shape the future of ethical, transparent, and trustworthy AI while remaining a world leader in AI innovation.
Risk Analysis: State agencies will perform a joint risk analysis of potential threats and vulnerabilities related to GenAI’s impact on critical energy infrastructure.
Procurement Guidelines: Agencies will issue guidelines for public sector procurement, use, and training related to GenAI, emphasizing safety and ethics.
CPPA’s draft bill on automated decision-making technology (ADMT) regulations: Focuses on businesses’ use of automated decision making for employees and consumers, reinforcing rights to opt out of and access information about businesses' use of automated tools and requiring businesses to conduct risk assessments in certain situations.
Notice of Use: Businesses that use ADMT to provide employees and applicants with advance notice that informs them about the use of the technology including profiling, the decision-making process and outputs, and their rights to opt out of and access information about the use of such technology.
Risk Assessments: The proposed regulation and the risk assessment regulation require businesses to conduct risk assessments where the "processing of consumers' personal information presents significant risk to consumers' privacy."
Alignment with other regulations: Restrictions to age specific behavioral profiling, option for opt-out and verifiable parental consent required under COPPA.
Differences
Focus:
Utah emphasizes consumer protection and accountability for businesses.
California prioritizes ethical, transparent, and trustworthy AI while maintaining its AI leadership.
Regulatory Scope:
Utah expands existing laws to encompass AI usage.
California’s executive order focuses on evaluation, deployment, and risk analysis.
Testing Requirements:
Utah’s approach is relatively light-touch, relying on existing laws.
California’s proposed legislation (Senate Bill 1047) requires mandatory testing for wide-reaching AI products before public release.
Other US States on AI
While federal legislation remains limited, several states have taken proactive steps to address the potential harms and benefits of artificial intelligence. Here’s a snapshot of their efforts:
Connecticut: Connecticut is among the states actively considering AI legislation. While specific bills are still in draft form, they share the goal of balancing stronger protections for constituents with fostering AI innovation.
Illinois: Algorithmic Transparency in Hiring Act: Illinois passed this law, which restricts the use of AI in hiring processes. It aims to prevent discriminatory practices and promote transparency.
Texas: Consumer Privacy Laws: Texas recently passed comprehensive consumer privacy legislation that includes provisions regulating AI. This mirrors similar laws in California, Colorado, Connecticut, and Virginia.
Other States:
Alabama, Colorado, Indiana, Louisiana, Mississippi, and New York City have enacted laws limiting AI use or prohibiting discriminatory algorithmic decision-making.
Montana, Oregon, and Tennessee have also passed consumer privacy laws with AI-related provisions.
Conclusion
These state-level efforts collectively contribute to shaping responsible AI adoption while safeguarding citizens’ rights and fostering technological progress. In summary, Utah’s pragmatic approach complements California’s comprehensive strategy, and both states contribute to shaping responsible AI adoption. Other states are taking the lead in addressing AI regulation, recognizing the need to protect citizens while fostering technological advancement. Their efforts vary, but the common thread is a commitment to good governance and thoughtful balance between protection and innovation.
References
Friedler, S., Venkatasubramanian, S., & Engler, A. (2023). How California and other states are tackling AI legislation. Brookings
Brennan Center for Justice. (2023)States Take the Lead on Regulating Artificial Intelligence.
EPIC – Electronic Privacy Information Center. (2023)The State of State AI Laws: 2023.
VentureBeat. (2022)AI regulation: A state-by-state roundup of AI bills.
Utah Approves AI Policy Act — What You Need to Know (cdomagazine.tech)
California lawmakers take on AI regulations as Congress dawdles : NPR
Draft California AI Regulations Become One Step Closer to Reality: An Analysis of Requirements on the Horizon | Wilson Sonsini (wsgr.com)
How California Lawmakers Are Tackling AI in 2024 (govtech.com)
California Agency Releases First Draft of AI Regulations (shrm.org)
What’s next for AI regulation in 2024? | MIT Technology Review
Comments