Researchers find healthcare AI’s sustainability efforts miss the bigger picture

Healthcare AI’s sustainability efforts are addressing pieces of the problem rather than the whole, a new research review finds. After examining 47 studies that considered sustainability, the authors call for environmental checks spanning manufacture, use and recycling. Their argument meets a practical tension in cancer care: clinicians see patient benefits that are difficult to refuse.

The paper, “Green artificial intelligence in health applications,” was published in Artificial Intelligence in Medicine in 2026. Co-author Alok Mishra, a professor in data management and software engineering at the Norwegian University of Science and Technology, argues that environmental consequences deserve equal consideration with privacy, bias, fairness and transparency—not treatment as a separate concern.

An efficient model is not the whole footprint

The researchers systematically reviewed healthcare-AI studies in which sustainability was already a consideration. They described the results as “siloed”: well-intentioned initiatives that did not consistently account for the bigger picture. That finding concerns how sustainability was approached within the selected studies, rather than a survey of every AI system used in medicine.

Their proposed alternative is a life-cycle assessment—an examination of environmental effects from manufacture through operation and eventual recycling. Under that approach, an efficient AI model or an energy-optimized data center is only one part of the assessment. Nor does using the technology for a beneficial purpose, by itself, establish that the application is environmentally sustainable.

Mishra compares environmental disclosure with an appliance’s energy rating or an airline showing passengers carbon-emissions information. But he acknowledges a barrier: the AI industry is often reluctant to disclose energy use and data storage. The assessment he advocates therefore asks for a wider view while depending on information that providers may not readily share.

Cancer care makes the benefits concrete

Åsmund Flobak, an oncologist at St. Olav’s Hospital’s Cancer Clinic, describes a different starting point: the patient in front of him. In his research, patient cells are grown as “living biopsies” and exposed to drugs. AI image analysis helps assess which cells die and which survive, supporting investigation of how treatments affect those patients’ cells.

At the clinic, Flobak says AI also assists in drawing the areas to irradiate or spare during cancer radiotherapy. Staff have begun using automatically generated consultation notes. Asked whether environmental impact is considered when introducing the technology, he says it is not actively discussed and should be addressed at the policy level. He points to limiting radiation doses to healthy tissue as a benefit that is difficult to turn down.

In my daily life as a doctor, I prioritize every benefit I can give my patients, within the guidelines and regulations I work under.

Åsmund Flobak, oncologist at St. Olav’s Hospital

Environmental costs can return as health pressures

Mishra’s concern extends beyond the clinic. He says AI’s data centers consume substantial electricity and fresh water, and argues that environmental consequences cannot remain only a political question. He points to landslides, melting glaciers, flash floods and heat waves as pressures on health systems. His argument is that environmental conditions are part of healthcare’s operating reality, not a concern wholly outside medicine.

That does not amount to a call to abandon medical AI. Mishra says researchers want it used sustainably so it can deliver more benefits to society. Flobak likewise works within guidelines and regulations rather than rejecting oversight. Their emphasis differs: one wants environmental responsibility built into development and application; the other looks to policy to set the boundaries within which doctors pursue patient benefits.