AI news & views roundup: AI-burdened nurses, AI-stealthy doctors, patient AI collaborative, more

Nurses are struggling to accept workplace assistance from AI.

Not all, of course, but enough of them that those who appear to be resisting the relentless march of progress catch the public’s eye. Scientific American looks at the burden carried by these essential healthcare workers behind the scenes in a reported article posted Feb. 17. The author is veteran reporter and NYU journalism professor Hilke Schellmann. Among her compelling observations: 

• Nurses are increasingly wary of unvalidated tools. “This friction has moved from the ward into the streets,” Schellmann points out. “In the past two years nurses in California and New York City have staged demonstrations to draw attention to unregulated algorithmic tools entering the healthcare system, arguing that while hospitals invest in AI the bedside remains dangerously short-staffed.”
   
• Sepsis prediction has become a cautionary case. “Hospitals across the U.S. widely adopted information health technology company Epic’s sepsis-prediction algorithm. Later evaluations found it substantially less accurate than marketed,” Schellmann reminds. To be sure, the company went on to release a new and improved version. Still, through the protracted episode, “nurses saw how an imperfect product could become policy—and then become their problem.”
   
• Nurses Schellmann interviewed for her article consistently emphasized that they are not opposed to technology in the hospital. “Many said they welcome tools that are carefully validated and demonstrably improve care,” Schellmann reports. “What has made them wary, they argue, is the rapid rollout of heavily marketed AI models whose performance in real-world settings falls short of promises.”
   
  • Read the whole thing. 

Dear hospital leadership: You can’t stop clinicians from bringing their own AI and using it in ‘the shadows.’

But you can get them to be upfront about it and thus more responsible with it. A physician and evident AI aficionado fleshes out the viewpoint in commentary published by MedPage Today. The health sector still tends to treat shadow AI “as a behavior to scold rather than a reality to govern,” writes Henry Bair, MD, MBA, of Wills Eye Hospital in Philadelphia. He recommends three steps to try in lieu of reprimands, talkings-to or strongly worded memos. 

1. Make AI use speakable. Create non-punitive ways for clinicians to say what they’re actually doing—not as confessions but as anonymous surveys, focus groups, department meetings framed as show-and-tell sessions and so on. “The goal is to learn which workflows clinicians are trying to fix and where unsanctioned tools are filling real gaps,” Bair explains. 
    
2. Build a safe ‘front door’ for AI.Offer a governed, easy-to-reach generative AI interface that keeps data inside covered infrastructure, has appropriate agreements in place and logs basic usage, Bair suggests. “Include templates for common tasks—summaries, letters, patient-friendly explanations—so clinicians don’t need to reinvent the wheel,” he adds. “If the official tool is clunky or obviously weaker than public chatbots, shadow AI will persist.”
    
3. Retroactively govern what works. Treat shadow AI as a discovery engine, Bair urges. “If many clinicians are using an unapproved tool for the same task, that’s a signal,” he writes. “Study that workflow, evaluate safer alternatives, and, when a vetted option is in place, actively help people switch.” Early trials of ambient scribes, he notes, “suggest you can cut documentation time and burnout without sacrificing note quality when the tool is carefully implemented.”

Read the rest.

Here’s another doctor advising honesty about AI. And that’s for patients as well as clinicians.

“As an AI researcher, I believe that, when used appropriately, these large language models are the greatest tool for empowering patients since the invention of the internet,” writes Adam Rodman, MD, in The New York Times. “But they also carry new and barely understood risks, like degrading the relationship that patients have with doctors, or pulling people into spirals of anxiety as they pepper a chatbot with questions.” The internist and AI program director at Beth Israel Deaconess lays out some things he’d like his patients to do with AI before he sees them.

• Use AI to enhance doctor appointments. Before you head to the medical office, pull up your medical notes, remove all identifiable information and copy those notes into an AI tool, Rodman advises. Then give the model a current update of your health and relevant concerns and ask the chatbot to concisely summarize all this information. Finally, ask the chatbot: “Given this context about my health, please give me three questions I should ask my doctor about my cough during my upcoming visit.”
   
• Look out for sycophancy in your AI interlocuter. Large language models tend to suck up to their users. This can make a model’s medical guidance suspect if not downright dangerous, Rodman reminds. To avoid this, patients should be as specific as possible with their prompts. And whenever possible, feed the model notes from recent doctor visits. “I remind patients that if a conversation is increasing their distress or anxiety, sycophancy is probably at play and it’s time to talk to your physician directly,” Rodman writes. 
   
• Just as Dr. Bair advised immediately above, be honest about your use of AI. And again, that goes for doctors as well as patients. “One of the great ironies of medical care in 2026 is that both patients and doctors routinely talk to AI systems without disclosing it to each other,” Rodman notes. He cites a recent survey showing two-thirds of physicians used the technology for professional purposes in 2024. “That includes me—I now routinely use AI to search medical literature, I use AI scribes to take medical notes, and I use privacy-compliant models when I need a second opinion,” Rodman writes. “Being honest about our AI use, in my experience, strengthens trust between doctors and patients and will hopefully lead to better care.”

Also worthwhile:

• Groundbreaking initiative establishes a standard for responsible patient-facing AI (AI Care Standard/PatientAI Collaborative) 
   
• What on earth is wrong with intelligent medicine? (Forbes)
   
• Dear Doctors: Can’t AI help reduce fatal medical errors? (UCLA Health)

From HealthExec’s sibling news outlets:

• AI model could improve early detection of deadly pregnancy complication (Radiology Business) 
   
• AI-powered MRI evaluations predict STEMI outcomes better than existing risk scores (Cardiovascular Business)