Nurses' unions in Michigan and New York City are winning contract language that gives them direct oversight of how artificial intelligence tools are deployed in their hospitals. The concessions come as AI adoption in healthcare accelerates, and front-line clinicians push back against technology they say could undermine patient safety and nursing judgment.
Nurses rally for AI protections
At Munson Medical Center in Traverse City, Michigan, the nurses' union organized a rally earlier this year to press for new contract terms. Union president Laura Nilsson listed demands that included higher pay, better staffing, and specific regulations around AI. "We understand technology is coming," Nilsson told the crowd. "We generally want it to be good for us and our patients, but right now, there [are] a lot of unknowns. What we don't want technology to do is replace nursing judgment. We don't want it to take jobs."
A nurses' union in New York City with more than 10,000 members secured similar AI protections in its contract this winter. The unions argue that even tools that automate certain tasks cannot replicate a nurse's physical presence and clinical assessment.
Technology handed to nurses without input
James Walker, a critical care nurse at Munson Medical Center and a member of the union bargaining committee, described how automated data collection can go wrong. "Say we're checking blood pressure and … the blood pressure cuff isn't on the patient appropriately," Walker said. "And so that number is just pulled into the system without verifying whether it's accurate … that value … it could potentially lead to a wrong decision."
Max Topaz, a professor of nursing at Columbia University, said that nurses are often left out of the design and implementation process. "Frequently, nurses are not involved in designing those technologies and thinking about how these technologies are going to be integrated into practice," Topaz said.
Error rates in AI models signal caution
In a recent study by researchers from Stanford and Harvard, even top-performing AI models from Anthropic and Google produced 12 to 15 severely harmful errors per 100 cases. The worst-performing models made mistakes in 40 out of 100 cases. "This is exactly why nurses are concerned," Topaz said. "These technologies have to be vetted and validated, and we need to know what the error rates are."
Hospital sees AI as a tool to reduce burnout
Jennifer Standfest, chief nursing officer at Munson Medical Center, represented the hospital at the bargaining table. "The growth of AI in healthcare really can be a positive thing. I mean, it is trying to support a reduction in some of the clerical tasks," Standfest said. Clerical tasks that cause burnout, she said, could be offset by AI. "It's a really important space, and we want to get it right."
Under the new contract, any decision by the hospital to buy new AI software will now include a voice for nurses in how it is rolled out.
Why this matters for healthcare
Healthcare professionals who work directly with patients are securing a formal role in evaluating AI before it reaches the bedside. The contract language sets a precedent that could spread to other hospitals and unions. For nurses and clinical leaders, the developments signal that AI adoption will require not just technical validation but also the active participation of those who will use the tools-and who will be held accountable for the outcomes.
Your membership also unlocks:
