As AI adoption accelerates, so do novel liability risks that do not neatly fit within traditional product liability and negligence frameworks developed for tangible products. Courts have been reluctant to treat standalone software as a product, particularly where it functions as an informational or advisory tool. However, plaintiffs increasingly argue that AI-driven health technologies, especially those embedded in devices or marketed as influencing clinical outcomes, should be treated as products subject to strict liability. Early case law shows courts grappling with how to apply existing product liability doctrines to technologies that continuously evolve through updates, ML, and real-world use, rather than remaining static at the point of sale.

Unlike conventional products, AI systems generate probabilistic outputs that complicate traditional warning paradigms underpinning design defect and failure-to-warn law. AI health product developers may face difficult questions about the scope and content of warnings: how to disclose limitations, known failure modes, or uncertainty in AI outputs – without undermining the utility of the technology or overstating risks.

There is little case law addressing duty of care, causation, and apportionment of liability when patients seek redress for purported injury related to use of AI tools. While cases involving non-AI clinical decision-making tools may be instructive, no clear framework has emerged in cases involving claims against AI developers. Causation is obscured by the fact that AI-driven health care decisions often involve multiple actors, including software developers, manufacturers, data providers, health care institutions, and clinicians. The opacity of AI tools further complicates liability: whether model bias in design or output caused actual injury; whether an incorrect output was foreseeable to the developer or end-user; whether such incorrect output resulted from the health care provider's use of the tool, the tool itself, or some combination; and whether the purported injury resulted from an incorrect output or other conduct.

Proactive risk assessment will be critical as AI technologies move from innovation to standard clinical practice. Making clear in marketing and training modules that AI tools are an adjunct to, and should not be used as a replacement for, independent medical decision-making could mitigate litigation risk. Contractual arrangements may also be useful defenses to liability. For example, in Sampson v. HeartWise Health Systems Corporation, the court held that because the cardiovascular disease-prevention software license made clear that the clinic retained final responsibility for interpreting scans and making diagnoses, the developer owed no duty of care as to those issues under a negligence theory.

As courts continue to confront these issues, companies deploying AI in life sciences and health care should anticipate heightened scrutiny of product characterization, design choices, warnings, and post-market performance.