The New York Times reported this week that "The Trump administration is accelerating efforts to make artificial intelligence an integral part of medical care in the United States, throwing the resources and support of the federal government into projects that deploy A.I. agents to diagnose and prescribe treatments to patients."
Depending on what news you've read recently, this could excite you or terrify you. I've previously written about my excitement about artificial intelligence (AI), but each day I am growing more worried that the way AI is being portrayed to the public is going to prevent patients from trying any drug, treatment, or tool that involves AI.
When CEOs of AI companies and developers are warning that the technology could "kill all humans," why should we expect anything but skepticism from patients and the clinicians who use AI tools?
It appears AI has a branding problem. To calm those fears and enable AI to support healthcare, the medical community needs to lead the way. We'll need to define what AI is, and what it is not; acknowledge AI will disrupt peoples' lives, sometimes in negative ways; admit there are risks to human welfare and create systems for accountability; and help patients sort the tools that pose a threat from those that help.
Put simply: healthcare executives and clinicians need to be vocal and lead.
What Is AI Anyway?
AI is everywhere. Or is it?
The algorithm that helps a radiologist flag a pulmonary nodule? We call that AI.
The ambient tool that drafts a note while I talk to a patient, so I am not charting at midnight? AI.
The system that flags a sepsis trajectory earlier than a busy resident? AI.
The frontier models now raising alarms about autonomous cyberattacks and engineered pathogens? Also AI.
It's one phrase, but these technological advancements should not be grouped in the same category. In fact, it's a major problem when we put them all in the same AI bucket.
To those of us who work with these tools, the distinctions are obvious. A narrow model trained to detect diabetic retinopathy has about as much in common with a self-directing frontier system as a blood pressure cuff has with a nuclear reactor.
But the public doesn't see it that way. To my patients, their families, voters, and maybe even to the regulators who ultimately will shape what we can and cannot do with AI, the distinctions are really blurry. Everything is now called "AI." That's a problem because it means they believe all these tools, many of which are life-saving or sustaining, are something to be feared, slowed, or completely disbanded.
AI can improve care, catch what tired humans miss, ease crushing workforce shortages, and maybe even cure cancer. But if patients, if society, rejects AI out of fear because they do not understand the nuance, the promise of the technology will never be realized.
The Growing Backlash Against AI
The backlash against AI has been swift and fierce. So fierce that the top companies have acknowledged it might be time to slow down progress.
An NBC poll published in early September found 70% of Americans are worried about AI. Young, old, wealthy, or poor, Americans have deep reservations about the technology. They are worried AI will kill them at worst or take their jobs at best.
It is no surprise, then, that clinicians are consistently more optimistic about AI than patients. A recent global index found roughly 79% of health professionals are hopeful it could improve outcomes. Only about 59% of patients agreed. More than half worried about losing the human touch in their care. A separate national study in JAMA Network Open found that patient trust in health systems to use AI responsibly was low.
Here is the part that should give healthcare professionals pause: the trust deficit had less to do with how much someone understood AI than with how much they trusted the healthcare system in the first place. Read that again. Comfort with medical AI is not a knowledge gap we can close with a pithy little explainer video.
It comes because people do not trust the system in general. Combine low trust in the healthcare community with zero trust in the people who run data centers and you can easily see the problem we are facing.
Distrust Threatens Investment
AI in medicine may hold promise, but Americans are forming their opinions of healthcare's future use of the technology far outside of exam rooms. They are listening to economists, politicians, AI company defectors, national security officials, and even their children's teachers who are banning it from classrooms. Even if none of it is about healthcare specifically, it all shares the label. And guilt by association is a powerful force.
The question I keep coming back to is this: has "AI" become such a broad and increasingly troubling brand that it could begin to thwart the implementation, investment, and adoption of some of its most useful medical applications?
I believe that is a real risk. If patients grow wary of anything labeled AI, they may decline tools that would genuinely help them, or lose confidence in the clinicians who use them. If that skepticism hardens, health systems may quietly slow rollouts or walk back projected savings. If the public mood sours enough to invite heavy-handed regulation aimed at frontier systems, narrow clinical tools could get swept up in the same net. Investors, reading all of this, may grow more cautious about the very applications most likely to help at the bedside.
There are legitimate concerns about AI, but we need to separate the tools that work from those that do not -- or those that may pose a threat.
That distinction exists in the technology. But it does not yet exist in Americans' minds.
So, What's the Prescription?
First, we need better words. "AI" is now nearly useless as a category. In medicine, we should name things plainly:
- Clinical decision support
- Ambient documentation
- Image analysis
We should also be honest about what each tool does, what it does not do, and who stays accountable when it is used. Patients will be reassured not by technical wizardry, but by human oversight and clear lines of responsibility.
Second, we should separate our story from Silicon Valley's. Medicine's use of these tools is relatively narrow, increasingly regulated (although gaps exist), and validated against outcomes. That is a fundamentally different enterprise than the race toward ever more capable general-purpose systems. We should stop letting the two be described in the same breath.
Third, we must keep earning trust the old-fashioned way. As the data show, it is the trust in us, not the sophistication of the tools, that patients are really evaluating.
The technology may be remarkable. But if we let "AI" become a scary word, some of the most useful applications in medicine could stall.