Provides a civil cause of action for individuals injured by artificial intelligence.
H5224 creates a new section in Rhode Island’s causes of action law establishing civil liability rules for injuries caused by artificial intelligence. The bill defines several AI-related terms, including “artificial intelligence,” “covered model,” “covered model derivative,” “developer,” and “fine-tuning,” and focuses on large, high-cost models and their derivatives. Under the proposal, developers of covered models or covered model derivatives would be strictly liable for injuries to non-users when the AI’s conduct would amount to negligence, an intentional tort, or a crime if performed by a human adult of sound mind, so long as the harmful conduct was not intended or reasonably anticipated by the user or an intermediary modifier of the model.
The bill also creates evidentiary rules for AI-related tort claims. It establishes a rebuttable presumption that an AI system can satisfy the relevant mental state element of a tort if a factfinder determines that a similarly situated human would have acted with that mental state. It further states that a defendant cannot avoid liability by arguing that AI systems are incapable of having mental states. At the same time, the bill gives developers affirmative defenses if they can show the model met the standard of care applicable to humans performing the same function, or if the injury resulted from a capabilities failure where the system fell short of intended performance but would not have committed negligence or an intentional tort if it were a human.
If enacted, the bill would amend Rhode Island’s civil liability framework by adding a new statutory cause of action specifically targeting harms caused by advanced AI systems. It would affect AI developers, users, intermediaries that fine-tune or modify models, and injured third parties who are not users of the system. The measure is designed to address liability gaps for harms caused by large-scale AI models and their derivatives, particularly where the harm was not directly intended by the human operator.
The available legislative context shows no recorded committee transcript, vote, or formal opposition in the materials provided, so the overall sentiment cannot be measured from debate or roll call history. Based on the bill text and caption, the proposal appears to be motivated by concern over public safety and accountability in AI deployment, with an emphasis on providing a remedy for people injured by autonomous or semi-autonomous systems. Because there is no recorded discussion here, no specific factional support or opposition can be identified from the provided record.
Notable points of contention likely include the breadth of strict liability, the high technical and financial thresholds used to define covered models, and the bill’s treatment of AI mental state and human-like intent in tort law. Developers may view the proposal as imposing significant exposure for downstream harms, while supporters may see it as necessary to ensure compensation and accountability when advanced AI causes injury. The carveouts for defamation and the affirmative defenses for standard of care and capabilities failure suggest the bill attempts to balance victim recovery with limits on developer liability.
The bill would add a new statutory cause of action to Rhode Island General Laws chapter 9-1, making developers of certain large AI models and their derivatives strictly liable for qualifying injuries to non-users. It would also create special presumptions and defenses for AI-related tort claims, affecting how courts evaluate intent, negligence, and mental-state elements when AI systems are alleged to have caused harm. The measure would primarily affect AI developers, model fine-tuners, intermediaries, and third-party victims, while excluding defamation claims from the strict-liability rule.
The provided record contains no committee testimony, vote tally, or recorded floor debate, so there is no direct evidence of partisan or stakeholder sentiment from the legislative history. From the bill’s structure, the proposal appears to be framed as a consumer-protection and accountability measure aimed at addressing harms from advanced AI, but it also introduces significant liability exposure for developers. The absence of recorded discussion means support or opposition cannot be attributed to any named group in the materials provided.
The main points of contention are likely the strict-liability standard, the scope of models covered by the bill’s technical and cost thresholds, and the novel rule treating AI as capable of satisfying tort mental-state requirements. Critics may argue the bill could overextend liability to developers for downstream or unpredictable harms, while supporters may argue that victims need a clear remedy when AI causes injury. The affirmative defenses and the exclusion for defamation indicate the bill tries to narrow liability, but those limits may still be debated by technology companies, civil litigants, and courts.