Artificial intelligence (AI) has rapidly reshaped cardiovascular imaging, particularly echocardiography, where machine learning algorithms now assist in image acquisition, chamber quantification, valvular assessment, strain analysis, and diagnostic interpretation. While these technologies promise enhanced efficiency and diagnostic accuracy, they also introduce an unresolved legal dilemma: When AI contributes to a missed diagnosis, who bears responsibility? Traditional malpractice principles designate the interpreting cardiologist as the primary legal decision-maker. Yet these doctrines evolved when diagnostic tools were passive instruments rather than autonomous or semi-autonomous analytical systems. This article argues that AI has fundamentally disrupted the legal assumptions underlying modern medical malpractice. Traditional tort doctrines (including negligence, product liability, corporate negligence, and the learned intermediary doctrine) must therefore be reassessed to allocate responsibility more equitably among physicians, hospitals, and AI developers while preserving physician accountability for independent clinical judgment.
Artificial intelligence is no longer a theoretical innovation in cardiovascular medicine. FDA-authorized algorithms now assist clinicians in evaluating left ventricular ejection fraction, detecting valvular heart disease, identifying hypertrophic cardiomyopathy, recognizing subtle wall-motion abnormalities, and automating quantitative echocardiographic measurements. As these technologies become increasingly integrated into routine clinical practice, physicians are expected to incorporate AI-generated recommendations into their diagnostic reasoning. Despite these advances, AI remains imperfect. A missed diagnosis of severe aortic stenosis, infective endocarditis, cardiac amyloidosis, hypertrophic cardiomyopathy, or left ventricular thrombus may result in catastrophic consequences, including stroke, heart failure, or death. When such an error follows an AI-assisted interpretation, litigation is almost inevitable. The more difficult question is whether existing malpractice law is adequately equipped to fairly assign legal responsibility.
Historically, medical malpractice has rested upon four familiar elements: duty, breach, causation, and damages. Physicians owe patients a duty to exercise the degree of care ordinarily expected of a reasonably prudent specialist under similar circumstances. Because the interpreting cardiologist signs the final echocardiographic report, current malpractice law generally assigns primary responsibility to the physician regardless of whether AI contributed to the diagnostic process. This framework developed when medical devices merely enhanced physician judgment. AI challenges that historical premise by performing sophisticated analytical tasks that increasingly resemble expert clinical reasoning. The legal question is therefore no longer simply whether the physician exercised reasonable care, but whether traditional negligence doctrine adequately reflects a diagnostic process now shared between human clinicians and intelligent software.
The complexity of technology-related liability is illustrated by two landmark Supreme Court decisions. In Riegel v. Medtronic, Inc., decided in 2008, the Supreme Court held that certain state-law claims involving FDA-approved Class III medical devices may be preempted by federal law. Conversely, in Wyeth v. Levine, decided in 2009, the court concluded that FDA approval does not automatically shield manufacturers from state tort liability. Although neither case involved artificial intelligence, both demonstrate that regulatory approval alone does not determine civil liability. As AI-enabled echocardiographic software increasingly receives FDA authorization, courts will inevitably confront analogous questions regarding whether software developers should share responsibility when algorithmic errors contribute to patient harm.
Real physician voices, twice a week
Free, and one click to unsubscribe.
Equally important is the learned intermediary doctrine, which provides that manufacturers generally satisfy their duty to warn by informing the treating physician rather than the patient directly. This doctrine assumes that physicians independently evaluate all available clinical information before making treatment decisions. That assumption becomes increasingly difficult to sustain when highly sophisticated AI systems generate diagnostic conclusions through complex neural networks that physicians cannot fully interpret, replicate, or explain. Requiring physicians to independently verify every AI-generated conclusion may substantially diminish the practical value of AI, while permitting unquestioning reliance would undermine professional accountability. Existing legal doctrine offers no satisfactory resolution to this tension.
The admissibility of AI-generated evidence may also present future challenges under Daubert v. Merrell Dow Pharmaceuticals, Inc., the 1993 decision that established the framework for determining the reliability of expert scientific evidence in federal courts. As AI increasingly influences echocardiographic interpretation, courts may eventually be required to determine whether algorithmic analyses themselves satisfy Daubert’s standards of scientific reliability or whether only physician testimony interpreting AI outputs should be admissible. This question extends beyond malpractice litigation and reaches the broader issue of whether artificial intelligence may function as an evidentiary expert in future courtroom proceedings.
Hospitals likewise cannot escape scrutiny. Modern malpractice jurisprudence recognizes corporate negligence, a doctrine firmly established in Darling v. Charleston Community Memorial Hospital, the 1965 Illinois decision that imposes an independent duty upon health care institutions to maintain safe systems of care. Applied to artificial intelligence, this doctrine raises important questions regarding hospital responsibility for vendor selection, algorithm validation, physician education, cybersecurity, software updates, quality assurance, and post-deployment surveillance. If a health care institution deploys inadequately validated AI technology or fails to identify declining diagnostic performance despite available evidence, institutional negligence may contribute to patient injury as significantly as any individual physician’s interpretation.
AI developers similarly may face liability under traditional product liability principles. Unlike conventional imaging equipment, however, AI software continuously evolves through algorithm refinement and software updates. Plaintiffs may argue that developers negligently designed the algorithm, inadequately warned users regarding known limitations, failed to update software after recognizing safety concerns, or overstated diagnostic performance during marketing. Manufacturers will likely respond that their products function only as clinical decision-support tools and that physicians retain ultimate authority over diagnosis and treatment. Future litigation will therefore require courts to distinguish software malfunction from physician judgment while determining whether AI developers should assume responsibilities comparable to those imposed upon manufacturers of sophisticated medical devices.
Importantly, protecting cardiologists does not require eliminating physician accountability. Rather, courts should distinguish between reasonable reliance upon a properly validated AI system and blind substitution of algorithmic output for independent clinical judgment. A cardiologist who carefully reviews echocardiographic images, integrates the patient’s clinical presentation, recognizes inconsistencies between AI findings and clinical evidence, and exercises independent medical judgment should not automatically bear exclusive responsibility when an otherwise properly functioning AI system contributes to a missed diagnosis. Conversely, physicians who merely approve AI-generated reports without meaningful review should remain subject to traditional negligence principles. The legal standard should therefore focus not on whether AI participated in the diagnostic process, but on whether the physician’s reliance upon AI remained objectively reasonable under the circumstances.
Artificial intelligence has fundamentally altered the relationship between physician, technology, and patient. For more than a century, malpractice law has been predicated upon the assumption that physicians exercise complete control over diagnostic reasoning while medical devices function merely as passive extensions of clinical judgment. Artificial intelligence challenges that assumption for the first time in modern medicine by introducing technology capable of performing sophisticated diagnostic analysis. If courts continue to apply legal doctrines developed for conventional diagnostic instruments to autonomous or semi-autonomous AI systems, they risk assigning liability according to historical tradition rather than technological reality. The next generation of medical malpractice law should therefore evolve from physician-centered liability toward a framework of comparative responsibility that recognizes the intertwined legal obligations of cardiologists, health care institutions, and AI developers. Such an approach would better reflect contemporary cardiovascular practice while preserving patient safety, professional accountability, and responsible innovation in artificial intelligence.
This essay is cited in the KevinMD records on artificial intelligence and heart disease.
Teddy A. Teddy is an internal medicine resident.
