A 8833: Establishes understanding artificial intelligence responsibility act requiring developers of covered models to be strictly liable for certain injuries.

  
  STATE OF NEW YORK ________________________________________________________________________ 8833 2025-2026 Regular Sessions  IN ASSEMBLY June 9, 2025 ___________ Introduced by M. of A. BORES -- read once and referred to the Committee on Science and Technology AN ACT to amend the general business law, in relation to establishing the understanding artificial intelligence act The People of the State of New York, represented in Senate and Assem- bly, do enact as follows: 1 Section 1. The general business law is amended by adding a new article 2 45-A to read as follows: 3 ARTICLE 45-A 4 UNDERSTANDING ARTIFICIAL INTELLIGENCE 5 ACT 6 Section 1510. Short title. 7 1511. Definitions. 8 1512. Artificial intelligence; liability for injuries. 9 § 1510. Short title. This act shall be known and may be cited as the 10 "understanding artificial intelligence act". 11 § 1511. Definitions. For purposes of this article, the following terms 12 shall have the following meanings: 13 1. "Artificial intelligence" means an engineered or machine-based 14 system that varies in its level of autonomy and that can, for explicit 15 or implicit objectives, infer from the input it receives how to generate 16 outputs that can influence physical or virtual environments. 17 2. "Covered model" means an artificial intelligence model trained 18 using a quantity of computing power greater than 1026 integer or float- 19 ing-point operations, the cost of which exceeds one hundred million 20 dollars when calculated using the average market prices of cloud compute 21 at the start of training as reasonably assessed by the developer. 22 3. "Developer" means a person that performs the initial training of a 23 covered model by training a model using a sufficient quantity of comput- 24 ing power and cost. EXPLANATION--Matter in italics (underscored) is new; matter in brackets [ ] is old law to be omitted. LBD02147-04-5 
 A. 8833 2 1 4. "Fine-tuning" means adjusting the model weights of a trained 2 covered model by exposing it to additional data. 3 § 1512. Artificial intelligence; liability for injuries. 1. Except 4 with respect to any causes of action for defamation, developers of 5 covered models shall be strictly liable, regardless of the degree of 6 care they exercised, for all injuries to a non-user of the covered model 7 that satisfy the actual harm element of an ordinary negligence claim if: 8 (a) those injuries are factually and proximately caused by a covered 9 model that engages in conduct that, if undertaken by an adult human of 10 sound mind, would satisfy the elements of negligence or any intentional 11 tort or crime; and 12 (b) that conduct was neither intended nor could have been reasonably 13 anticipated by: (i) the user of the model; or (ii) any intermediary that 14 fine-tuned, scaffolded, or otherwise modified the model. 15 2. (a) For the purposes of this section, for any torts for which the 16 mental state of the alleged tortfeasor is relevant to elements of the 17 tort, there shall be a rebuttable presumption that the artificial intel- 18 ligence system satisfies the relevant mental state if the finder of fact 19 determines by a preponderance of the evidence that, if a natural person 20 under similar circumstances to the artificial intelligence system took 21 actions similar to those taken by the artificial intelligence system, 22 that natural person would have acted with the relevant mental state. 23 (b) Unless the court determines that the presumption established in 24 paragraph (a) of this subdivision is not applicable, if the party 25 against whom the presumption is invoked presents evidence tending to 26 rebut the presumption established in paragraph (a) of this subdivision, 27 the court shall instruct the finder of fact to find that the presumed 28 facts exist unless the finder of fact is persuaded that the presumed 29 facts do not exist. 30 (c) For the purposes of this section, it shall not be a defense that 31 artificial intelligence systems are incapable of having mental states. 32 3. (a) It shall be an affirmative defense to strict liability if the 33 developer establishes that the covered model satisfied the standard of 34 care applicable to humans who perform the same function that the covered 35 model was engaged in performing when its conduct allegedly caused the 36 plaintiff's injury. 37 (b) It shall be an affirmative defense to strict liability if the 38 developer establishes that the injuries to a non-user as described in 39 this section were a result of a capabilities failure, in which a covered 40 model falls short of performing the intended or reasonably anticipated 41 performance of the user, but the conduct of the system would not satisfy 42 the elements of negligence or any intentional tort or crime if engaged 43 in by an adult human of sound mind. 44 § 2. This act shall take effect immediately.