The Nurses Union Breaking Ranks to Fight AI in Hospitals

Chris Mills Rodrigo is a fellow at Tech Policy Press.

Most of the protest messages that nurses picketing NewYork-Presbyterian Hospital displayed this winter struck familiar labor themes — that their work is important, they deserve to be fairly compensated and their bosses are greedy.

But many of the nurses on strike with the New York State Nurses Association also brandished more novel signs, sporting messages like “Ask for a real nurse” and “Pay me like one of your AI robots.”

National Nurses United, the parent union of the New York group, has distinguished itself as one of the earliest and most forceful labor voices against the adoption of artificial intelligence in their members’ workplaces.

The nurses union is not taking a fully rejectionist stance — they don’t foreclose the possibility that AI could help their members — but has been notably outspoken about how it has been deployed so far compared to most other United States unions.

Many labor groups have adopted a far less adversarial approach to AI integration in the workplaces they represent, calling for worker involvement in deployment but not rejecting the technology.

Cathy Kennedy, a president of NNU and a recently retired nurse of 46 years, credits the union’s early AI opposition to its long-standing fight against earlier algorithmic tools presented as potential solutions to nursing shortages.

“Our eyebrows started raising [about] the introduction of AI in the early 2000s,” Kennedy told Tech Policy Press in an interview.

Kaiser Permanente hospitals in California, where Kennedy worked, introduced its own Electronic Health Record (EHR) system in 2004 in partnership with the firm Epic Systems.

EHRs are digital versions of patient medical histories, designed to make it easier for medical professionals to keep track of people’s conditions over time. For nurses in hospitals, these tools can be very important during patient handoffs between shifts.

The issue, for Kennedy and other NNU members, arises when EHRs replace human-to-human communication between nurses and other staff. The systems, deployed in hospitals across America, analyze the information submitted into them to make some decisions that staff would otherwise make, like discharge times or staffing needs.

So while EHRs were not branded as being AI-powered when they were introduced, their experience with them has given NNU members years of experience pushing back against the introduction of algorithms into their workplaces.

“AI cannot do what a registered nurse can do in looking at a situation critically, individualizing the care,” Kennedy said.

Hospitals have increasingly integrated AI technologies into their health record systems over the last few years, primarily to make predictions about health trajectories or risks for patients. Epic, which in addition to running EHR systems at many hospitals including Kaiser also powers the ubiquitous patient portal MyChart, says on its website that it “moved quickly to embed [AI] throughout our software.”

NNU has strongly pushed back on this process, arguing that AI systems generate false alarms, miss things human nurses wouldn’t and have baked-in biases that will make care outcomes worse for patients from underserved communities.

A survey of members conducted by the union found that significant portions of nurses disagreed with the content in AI-generated patient risk assessments and hand-offs for the next nursing shift.

The union has been particularly outspoken against efforts to paint AI as a solution to nursing shortages in hospitals.

Department of Health and Human Services Secretary Robert F. Kennedy Jr. and Centers for Medicare & Medicaid Services Administrator Mehmet Oz have both touted the potential for AI-powered nurses to provide health care to areas without enough doctors and nurses.

Kennedy said during his confirmation hearing that “you cannot distinguish” AI nurses developed by one clinic from a human being and that they can provide diagnoses “as good as any doctor.”

The tech industry appears to see potential in that idea as well. Nvidia, for example, launched an AI-powered virtual nurse project with Hippocratic AI where each agent only costs $9 an hour to operate. Registered nurses tend to make roughly $40 an hour, according to the U.S. Bureau of Labor Statistics.

Cathy Kennedy, the nurses union president, is not convinced by the potential of AI-nurses to replace human care.

“When the lights go out, who's going to be there for you? It's not going to be your virtual nurse,” she said. “It's going to be a human being.”

Instead of introducing new technology aimed at making nurses’ jobs easier, or even replacing nurses, Kennedy and the NNU want there to simply be more nurses in hospitals. The union has been advocating for federal legislation mandating safe staffing ratios, pointing to research suggesting lower patient-to-nurse ratios reduce mortality rates and help nurses avoid burning out of the profession.

NNU’s position on AI — or at least how it’s being deployed now — stands apart from other US unions not only in being more critical, but also in how early it was adopted. While others have joined in on speaking out against growing AI use, NNU was one of the first.

“Our organization is right up front because [of] our values of caring, compassion and community,” Kennedy said.