Bill > SB3590

This bill, titled the Artificial Intelligence Product Liability Act, establishes new rules for holding developers and deployers of artificial intelligence (AI) systems accountable for harm caused by their products. It defines "artificial intelligence" broadly and specifies that a "product" under this act refers to a "high-impact artificial intelligence system" (AI that significantly affects legal or important aspects of life, categorizes people by protected traits, manages critical infrastructure, is used in vehicles or medical devices, mimics human interaction, or poses serious risks) or a "generative artificial intelligence system" (AI that creates new content). The bill outlines that developers can be held liable if they fail to exercise reasonable care in the design of the AI, in providing adequate instructions or warnings, or in upholding an express warranty, provided these failures are the direct cause of harm. A "deployer," which is someone who uses or operates an AI system, can be held liable as a developer if they make substantial changes to the AI or intentionally misuse it against an express warranty, and this misuse directly causes harm. The act also sets forth specific requirements for developers to avoid liability, including conducting thorough testing and risk assessments, disclosing risks, and maintaining an "artificial intelligence data sheet" detailing the AI's training data, intended uses, and risk mitigation strategies, referencing guidelines from the National Institute of Standards and Technology (NIST). The bill supplements existing laws and applies to actions filed on or after its effective date, with certain exceptions for scientific research.

Government Affairs

https://www.ilga.gov/Legislation/BillStatus?DocNum=3590&GAID=18&DocTypeID=SB&SessionID=114&GA=104