The use of artificial intelligence: From clinical decision-making to courtroom strategy | Viewpoint

Artificial intelligence is rapidly transforming both healthcare delivery and medical malpractice litigation.

AI-driven clinical decision support tools empirically improve patient outcomes, while AI-powered legal analytics are reshaping case screening, discovery, and expert testimony. These developments create new opportunities and risks. This article examines AI’s emerging role for the clinician, and for the team of risk and legal professionals who defend them.

Understanding artificial intelligence is the first step to use it responsibly

Many clinicians, claim professionals and attorneys use the phrase “Artificial Intelligence” without a fundamental appreciation of what it means. But a reasonable understanding of AI is necessary to use it responsibly and to evaluate it critically when used by the plaintiffs’ bar.

A full recitation of what constitutes AI or its defined terms is beyond the scope of this article and the expertise of its author. But in simple terms, AI is statistical architecture overlapping data, that provides an output or makes decisions instead of following fixed rules. Architecture is the overall structural design of a model: how information flows through it, what components it uses, and how those components interact to transform input into output. This is proprietary work of the model that you can't see. Data is the world of information the model uses. For ChatGPT, the data is the Internet: the good, the bad, the true and untrue. AI models intended for professional use, such as Thomson Reuters’ Co Counsel or Epic’s integrated Copilots, change the data source and hence the reliability of the model.

For both clinicians and attorneys the lesson is quite clear: whatever the AI model, verification of output together with your own medical and legal judgment remains important. Misinformation, bias, and “hallucinations,” when a model produces confident but incorrect information, remain significant obstacles. Because AI outputs are probabilistic, clinicians and attorneys must verify recommendations against authoritative sources. AI should be treated as an assistant whose work requires supervision, not as an independent decision-maker.

AI at the point of care

While some hospital systems have integrated Med-Gemini, a family of advanced AI models developed by Google, the year 2025 is most notable for Epic’s development of its own Copilots with a beta phased release throughout 2026. Some estimates show Epic holds as much as 42% of the Electronic Health Record (EHR) market, so these models are worth watching. Before this year, Microsoft/Nuances DAX Copilot was the most prominent and generally available AI tool in Epic. Its function was to use AI to listen to patient conversations and automatically draft clinical notes which are then available for review. Nuance stated in a 2024 data survey that users reported a 50% reduction in documentation time and a 70% reduction in burnout. This apparent success has generated more development.

In August 2025, Epic announced its own native AI tool “Art” designed to act as a “second set of eyes” for the clinician, asserting it will provide proactive support throughout patient care. One of Art’s more interesting features is to provide diagnostic assistance. Leveraging data from Epic’s extensive de-identified data set called “Cosmos,” Art suggests potential differential diagnosis, possible order sets and other treatment plans designed to aid clinicians in their decision-making process.

The risks of using these AI platforms, both for the clinician and the attorney who will later defend her, is a lack of training and understanding of how AI should be used. Some providers have expressed frustration with these models, worried that liability will be imposed if each suggestion is not ruled out.

In defending physicians and hospitals for over 30 years, I would suggest this is an old problem disguised as a new one: all clinicians have used medical literature or other professional resources such as UpToDate. We know that these are guidelines to consider alongside clinical judgment and the individual needs of the patient. For now, AI is no different. It's an occasion to look at suggestions and use your own clinical judgment to evaluate the AI output. That’s also how we guide our providers to advance through the rigor of providing testimony when in suit.

AI-driven litigation

Venture capital investment in litigation is its own frightening trend.

Now Venture capital is supporting “consumer protection firms” that produce AI platforms to identify potential litigation targets.

But AI’s introduction to professional resources also carries the promise of better legal work. Val.AI, a controlled, comparative study of multiple legal AI platforms, found that AI outperformed attorneys in document summarization, transcript analysis, and data extraction, matched attorneys in chronology generation, but underperformed in redlining and complex regulatory research.

Other studies, however, demonstrate that AI legal research tools produce hallucinated citations at rates between seventeen and thirty-three percent, necessitating manual verification of all authorities. The lesson for the defense attorney is the same as the clinician – we can improve our work with the use of AI, if it is coupled with verification and our own strategic review.

AI models are also reaching our experts. Commercial models are exploding online, claiming to be HIPAA compliant and inviting your expert to create medical chronologies and suggestions for standard of care and causation analysis. Defense experts should be immediately cautioned not to use such platforms without independent assessment and verification. Plaintiff experts should receive greater scrutiny at deposition and particularly in the courtroom. These models lack transparency regarding training data, architecture, and error rates. This opacity raises significant issues under the standards of Daubert, Frye and Rule 702. These reported cases are just starting to emerge and Defense counsel should be prepared to challenge AI-supported expert testimony on grounds of reliability, methodological transparency, and qualification.

Conclusion

Artificial intelligence is reshaping healthcare and medical malpractice litigation from the bedside to the courtroom. Those who understand AI’s strengths and limitations, verify its outputs, and integrate it strategically will be best positioned to deliver effective care and mount successful defenses in an AI-enabled legal landscape.

Brenda S. McClearn, Esq, is a partner at Hall Booth Smith P.C.’s Denver office.