AI ethics, regulation and use at the center of discussion at 2025 Dawn Gideon and Kevin Altomari Symposium

It’s nearly impossible to live life without interacting with artificial intelligence (AI). As with any novel technology, it has supporters and detractors as its use expands, especially as it gains popularity in fields like health care, where accuracy and privacy are top priorities.

The 2025 Dawn Gideon and Kevin Altomari Symposium brought together AI experts to discuss responsible AI use, the risks health care systems face when implementing new technology, and policymakers' efforts to regulate it as it evolves.

“You may be surprised at who all is involved in trying to regulate AI,” said Jolene Calla, Esq., vice president of finance and legal affairs for the Hospital and Healthcare System Association of Pennsylvania. “The World Health Organization is emphasizing a need for a framework that is responsible and consistent. They’ve set out some high-level principles but haven’t gone into the specifics.” In the U.S., the Food and Drug Administration is doing much of the AI regulation because the tools and devices using the technology - specifically, the safety and efficacy – fall under its jurisdiction. Calla says laws also govern AI use of data and mining.

Dr. Robert Bart is the chief medical information officer for UPMC and says a major project the system is currently undertaking is moving patient records to electronic health records, and says leaders in the system believe that AI can help not only improve patient care but also be used to spot trends or inefficiencies that can reduce inequities in care.

“Everything AI does is an algorithm based on programming and a data set it's learned from, so if the data is good, it can be equitable and fair,” he said, stressing that all AI has to be “taught” and, like any person, can develop biases or “bad habits” if not accounted for.

Like people who need to learn, AI needs the right people to influence it so it can make the best decisions and learn from its mistakes, said Tony Lukasavage, vice president of engineering at Innovu, LLC, a health care analytics company based in Pittsburgh.

Moderator Marty Raniowski (far left) asks panelists a question during the 2025 Dawn Gideon and Kevin Altomari Symposium. 

“Right now, AI is really good at summarizing and research, but it's not good at problem-solving,” he said. “At this stage, anyone who is using it for a new purpose beyond summaries or basic research needs to treat AI like an intern. Generally, it knows what it's doing, but it will need some help with more complex tasks.”

All three panelists agree that for expanded use in health care, systems and providers need to take caution to protect confidential information, but say it holds tremendous potential to help clinicians.

“I hope that AI can take patient-generated data and provide context to the caregivers,” said Bart. “That context and how a condition or comorbidity applies to a specific patient can help create a better model of care.” AI may cost a user on the front end, but it can save money in the long run, Bart adds, but warns administrators that it is a tool, not a trained clinician. “The biggest competition for hospitals isn’t other systems,” said Bart. “It’s Dr. Google. Someone feels ill, and they look up the symptoms, and if AI isn’t trained right, you could have people not seeking care or deciding to treat themselves with unintended consequences.”

Another issue posed by a technology that learns and could develop biases is what to do if the AI begins spreading misinformation. “It’s not an end-all, be-all,” said Lukasavage. “It’s like any other tool; it can do grunt work like cleaning data, but it needs context to learn, too.”