WeVote
Log In

Bill

Bill

SB 3590

AI PRODUCT LIABILITY ACT

104th Regular Session Introduced by Mary Edly-Allen

It creates a liability framework for AI products, holding developers and deployers to design, warn, and warrant standards with risk management and presumption defenses.

Rule 3-9(a) / Re-referred to Assignments
0
WeVote Research Nonpartisan
Bill Summary · SB 3590

Summary: Artificial Intelligence Product Liability Act (SB 3590, 104th Illinois General Assembly)

Overview

SB 3590 introduces the Artificial Intelligence Product Liability Act in Illinois. The bill creates a specialized framework for liability actions involving artificial intelligence (AI) systems, focusing on high-impact and generative AI. It assigns accountability to developers and deployers, sets standards for design, warnings, and warranties, and establishes evidentiary presumptions and defenses related to risk management, testing, and disclosure.

Main purpose and intent

  • To address harm arising from AI systems by clarifying when developers and deployers can be held liable.
  • To require reasonable care in design, instructions/warnings, and express warranties for AI products.
  • To incentivize robust risk management, transparency, and documentation for AI deployments.
  • To ensure that liability rules align with the unique use-cases and risks of high-impact AI.

Key provisions and changes

Definitions and scope

  • Defines AI concepts relevant to the act, including:
    • High-impact AI systems
    • Generative AI systems
    • Deployer (with exemptions for very small entities: fewer than 20 employees or fewer than 10,000 users)
    • Developer
    • Express warranties, material facts, and product releases
    • Harm, consequential decisions, synthetic relationships
  • Distinguishes product liability actions under this Act from other statutory or common-law claims, with exemptions for scientific research uses.

Developer accountability for harm (Section 15)

  • A developer is liable to a plaintiff only if the plaintiff proves, by a preponderance of the evidence: 1) Failure to exercise reasonable care in design that was a proximate cause of harm; and 2) Failure to provide adequate instructions or warnings that were a proximate cause; and 3) Failure to provide an express warranty, its product failed to conform, and the failure caused harm.
  • For defective design claims, the plaintiff must show:
    • The developer knew or reasonably should have known of the danger at the time the product left control,
    • Reasonable consideration of intended and foreseeable unintended uses,
    • Existence of a technologically feasible and practical alternative design that would have reduced risk without significantly impairing usefulness.
  • For failure-to-warn/instruction claims, the plaintiff must show: adequate warnings were not provided and were a proximate cause; warnings must be tailored to the danger and user knowledge; dangers known or obvious to users may limit liability, with a presumption of non-obviousness for users under 17.
  • For express warranty claims, liability arises if the plaintiff reasonably relied on a material warranty, the warranty proved false, and harm would not have occurred otherwise.

Deployer accountability for harm (Section 20)

  • A deployer is deemed liable as a developer if:
    • The deployer makes material and substantial changes to the product, or
    • The deployer intentionally misuses the product contrary to the express warranty, and that misuse caused harm.
  • Use that aligns with the product’s intended use is not considered misuse; if the product lacks an explicit intended use, the intended use is inferred from target market and distribution.
  • A deployer licensing a product cannot be held liable solely by virtue of ownership or use for violations committed by another.

Evidentiary presumptions and risk management (Sections 25, 15, 20, 24)

  • Rebuttable presumptions:
    • A product is not defective if the deployer conducted documented testing, evaluation, verification, validation, auditing to industry best practices (e.g., NIST AI RMF), mitigated risks, disclosed risks to deployers and consumers, and maintained an AI data sheet with specified contents (model context, data sources, risk management, red-teaming results, etc.).
    • A product is not defective if the deployer designed and implemented a risk-management policy meeting specified standards and is accessible to employees and the Attorney General.
  • Data sheet requirements (AI Data Sheet content):
    • Intended contexts and uses per NIST RMF guidance
    • Datasets used, sources, volume, proprietary status, and purpose
    • Risk management steps and red-teaming results
  • Age-appropriate safeguards for users under 17, including cognitive/emotional development assessments and age gating or content restrictions where appropriate.
  • Disclosure of risk information in terms and conditions to assist deployment decisions.

Applicability and enforcement

  • Applies to actions commenced on or after the act’s effective date; does not bar other AI-related liability under state law if not directly conflicting.
  • Exempts strictly peer-reviewed scientific research products.
  • Allows joint and several liability between developers and deployers; sets comparative negligence standards.

Who is affected

  • Developers of high-impact AI systems and generative AI systems.
  • Deployers using or operating AI systems, including those who modify or repurpose products (subject to material change and misuse provisions).
  • Entities with significant scale (non-excluded under the 20 employees/10,000 users thresholds) are more likely to be covered.
  • Consumers, businesses, and potentially public sector users affected by AI-driven harms, including property, personal, financial, reputational, or consequential harms.

Procedural and timeline aspects

  • The act sets forth standards for liability determinations, presumptions, and defenses that courts will apply in AI-related product liability cases.
  • It includes a framework for testing, documentation, and disclosure that, if met, creates a rebuttable presumption of non-defectiveness.
  • It emphasizes a comparative negligence framework, with potential joint and several liability and allocation aligned with each party’s fault.
  • Severability clause ensures provisions operate independently if any part is held invalid.

Note: The bill is introduced and subject to legislative debate, committee reviews, and potential amendments before any final passage.

Compiled from official sources — confirm details with the bill’s official record.

Sign in to ask a question.