AI Frontier Institute

Establishes understanding artificial intelligence responsibility act requiring developers of covered models to be strictly liable for certain injuries.

View primary source →

New York Assembly Bill 8833 (2025-2026), the "understanding artificial intelligence act," has not been enacted. It is in the Assembly Science and Technology Committee. If enacted, it would add Article 45-A to the General Business Law and take effect immediately. It would make developers of very large AI models strictly liable for certain harms to people who did not use the model. A "covered model" is one trained with more than 10^26 operations of computing power at a compute cost above $100 million (§ 1511). A "developer" is the person who performs the initial training of such a model (§ 1511). Except for defamation claims, developers would be strictly liable, regardless of the care they took, for injuries to a non-user that meet the actual-harm element of ordinary negligence (§ 1512(1)). Two conditions apply. First, the injury must be factually and proximately caused by a covered model engaging in conduct that would be negligence or an intentional tort or crime if done by an adult human of sound mind (§ 1512(1)(a)). Second, neither the user nor any intermediary that fine-tuned, scaffolded or otherwise modified the model could have intended or reasonably anticipated that conduct (§ 1512(1)(b)). For torts where mental state matters, there is a rebuttable presumption that the AI system had the relevant mental state. The presumption applies if a person acting similarly in similar circumstances would have had that mental state, as found by a preponderance of the evidence (§ 1512(2)(a)-(b)). It is not a defense that AI systems cannot have mental states (§ 1512(2)(c)). Developers would have two affirmative defenses. The first is showing the model met the standard of care that applies to humans performing the same function (§ 1512(3)(a)). The second is showing the injury resulted from a "capabilities failure" where the conduct would not be negligence or an intentional tort or crime if done by a human (§ 1512(3)(b)).

← Back to the tracker