The AI LEAD Act (Aligning Incentives for Leadership, Excellence, and Advancement in Development Act, S. 2937) is a bipartisan Senate bill introduced on September 29, 2025 in the 119th Congress by Sen. Dick Durbin (D-IL) and Sen. Josh Hawley (R-MO). It establishes a federal product liability framework for AI systems, treating AI as products subject to tort law, and has been referred to the Committee on the Judiciary. The bill's stated aims are to align incentives for safety and innovation, provide legal certainty, remove barriers to interstate commerce, and ensure US competitiveness.
Definitions
The bill defines its key terms as follows:
- Artificial intelligence system ("covered product"): any software, data system, or tool capable of making predictions, recommendations, actions, or decisions using machine learning, statistical/symbolic models, or other algorithmic methods.
- Developer: a person who designs, codes, produces, owns, or substantially modifies a covered product.
- Deployer: a person who uses or operates a covered product for personal/commercial use or for use by a third party.
- Harm: property damage, personal injury/death, financial/reputational injury, mental/psychological anguish, emotional distress, or distortion of behavior that would be "highly offensive to a reasonable person."
- Design: intended/known characteristics, data selection for training, training/testing/auditing/fine-tuning, and "unexpected skills or behaviors that appear in the covered product."
Key provisions
Developer and deployer liability (Title I)
Under Sec. 101, developer liability rests on four theories. Negligent design is the failure to exercise reasonable care in design, which requires showing that a reasonable alternative design existed unless the design is "manifestly unreasonable." Failure to warn covers inadequate instructions or warnings about foreseeable risks. Breach of express warranty applies where the product failed to conform to material promises. Strict liability applies where the product was in a defective condition unreasonably dangerous when used or misused in a reasonably foreseeable manner; under this theory the developer is liable regardless of whether they exercised all possible care.
Several provisions shape how defect is established. Circumstantial evidence can support an inference of defect without proof of a specific defect. Non-compliance with safety statutes or regulations is deemed defective, while compliance with regulations does not preclude a finding of defect. Risks that are "open and obvious" to ordinary users are excluded, but risks are presumed not to be open and obvious to users under 18. The developer's knowledge standard is whether they knew or "reasonably should have foreseen" the risk at the time of release.
Under Sec. 102, deployers are liable as developers only if they make a "substantial modification" to the product or intentionally misuse it contrary to its intended use. Uses intended by the developer do not constitute modification or misuse.
Unconscionable liability limitations (Title II)
Developers may not include contract terms that waive rights, limit forums, or unreasonably limit liability under the Act; such terms are unenforceable. The same prohibition applies to terms and conditions for deployers and end users.
Enforcement and preemption (Title III)
The bill creates a federal cause of action available to the Attorney General, state attorneys general, individuals, or class actions, with a 4-year statute of limitations that is tolled for minors and persons with legal disabilities. On preemption, the Act supersedes state law only where it conflicts, but states may enact stronger protections, making this a floor rather than a ceiling. Deployers can be held liable in place of developers who are outside jurisdiction or insolvent, and joint and several liability applies when both developer and deployer contributed to harm.
Foreign AI developer registration (Title IV)
Foreign developers must designate a US-based agent for service of process before deploying AI in the US. The Attorney General maintains a public registry, and failure to register prohibits a foreign developer from deploying AI in the US.
Effective date (Title V)
The Act applies to liability actions commenced after enactment, regardless of whether the harm occurred before enactment, providing for retroactive application to pre-existing harms.
Positioning within AI liability policy
The bill is the first major bipartisan federal AI liability bill, with the Durbin (D) and Hawley (R) co-sponsorship indicating cross-party support. It applies strict liability to AI developers, treating AI systems like consumer products rather than services or speech. The inclusion of "unexpected skills or behaviors" within the definition of "design" treats emergent capabilities as a design responsibility, a framing that connects to Emergent Misalignment and AI Autonomy Risk. The presumption that minors are unable to recognize AI risks addresses teen harm from AI chatbots, cited in the bill's findings, and the foreign developer registration provision creates a jurisdictional hook for foreign AI companies.
The product liability approach contrasts with the transparency-first approach favored by Dario Amodei and embodied in SB 53 and the RAISE Act. Rather than relying on disclosure, it creates direct financial incentives for safety through litigation risk. Its floor preemption, which allows states to go further, is consistent with the federalist approach described in Techno-Federalism. The bill does not address antitrust, export controls, or national security dimensions.
The strict liability provision for a "defective condition unreasonably dangerous" would apply to AI Autonomy Risk scenarios, since developers are liable even if they exercised all possible care, and the "unexpected skills or behaviors" language in the design definition encompasses Emergent Misalignment. The foreign developer registration provision connects to AI and Authoritarianism concerns about foreign AI systems, and the floor preemption with stronger state protections maps to the techno-federalism dynamic of states leading AI regulation.
The bill runs counter to Illinois SB 3444 — Artificial Intelligence Safety Act (Illinois Artificial Intelligence Safety Act, Feb 2026), a state-level conditional-immunity or safe-harbor bill that shields frontier developers from "critical harm" liability if they publish a safety protocol and transparency report. Where the AI LEAD Act expands developer exposure through strict liability, SB 3444 removes exposure through a safe harbor. Sen. Durbin (D-IL) co-sponsors AI LEAD federally while Sen. Cunningham (D-IL) sponsors SB 3444 in the state senate, producing opposing liability frameworks from the same state delegation at different levels of government. OpenAI publicly supports SB 3444, a departure from its posture on liability-expanding bills such as this one.
Provenance
This page summarizes the text of S. 2937 (119th Congress, introduced 2025-09-29).