Safety and alignment for clinical AI

A research lab in the Department of Radiology at Beth Israel Deaconess Medical Center. We test how clinical AI behaves when real patients are at stake.

Surgeons reading X-rays on a lightbox

Aligned with the patient in front of it.

BRIDGE is a team of physicians and AI researchers, including some of the most cited names in clinical AI, whose studies already make up a large part of the evidence on how language models behave in medicine. We test one thing above all: whether a model is aligned with the people it serves. That means patients who deserve the same care whoever they are, physicians who need tools they can trust, and health systems that need care to be better, more efficient and more equitable.

Eyal Klang
#1
in the world for large language model research
Eyal Klang, Co-Director. ScholarGPS ranking, across his career and over the past five years.

What we test for is alignment.

A patient's hand with a hospital wristband, held by another hand

Equity

Clinical AI should be sensitive to what truly changes care: the clinical facts and the evidence, not stigma, an unrelated label or the tone of a question. We measure this at scale and build solutions for consistent, equitable AI.

A hand radiograph with the joints a model has marked

Truthfulness

Clinical AI should know what is real and what the evidence supports, say so when it does not, and tell patients and clinicians the same. We study how models take in health misinformation and build frameworks for accurate health information and shared decisions.

Looking into a CT scanner, with a chest film inside the ring

Safety

Clinical AI should recognize an emergency, respond safely in a crisis and know when a person needs a clinician, wherever people use it. We test models in real settings, from triage to a late-night question at home, and build safeguards patients and health systems can rely on.

Who we are

Eyal Klang MD

Eyal Klang MD

Co-Director

Eyal is a radiologist in the Division of Abdominal Imaging and Interventions at Beth Israel Deaconess Medical Center. ScholarGPS ranks him first in the world in large language model research, and he has authored hundreds of peer-reviewed studies, including in Nature Medicine, The Lancet and JAMA.

Yiftach Barash MD

Yiftach Barash MD

Co-Director

Yiftach is an interventional radiologist at Beth Israel Deaconess Medical Center who completed his clinical fellowship at Harvard Medical School. An engineer by training with a background in industry, he holds a Harvard Medical School Shore Faculty Fellowship and applies machine learning and imaging to clinical decisions.

Mahmud Omar MD

Mahmud Omar MD

Head of Research

Mahmud is a family physician, a Lecturer at the Icahn School of Medicine at Mount Sinai, and the first AI research fellow at NEJM Group, where he develops technology for safe AI and health information. He has authored tens of peer-reviewed studies, including in Nature Medicine, The Lancet and The Lancet Digital Health.

Alon Gorenshtein MD

Alon Gorenshtein MD

Head of AI Engineering

Alon is a physician-scientist training in AI and neurology at Harvard Medical School and a postdoctoral fellow in the Epilepsy and Data Science Lab at Beth Israel Deaconess. He builds agentic AI for clinical neurophysiology, including INSPIRE for nerve conduction and EMG interpretation, and tools for clinical decision support.

What we have published

Selected recent papers by the BRIDGE team. For the full record, see PubMed or Google Scholar.

Follow to the story behind a paper.

Who we work with

Our network spans clinicians, data scientists, engineers and policy researchers across institutions. If you have an idea worth exploring, let's figure it out together.

Collaborators

Students

Work with us

Have a research idea, or want to collaborate on clinical AI evaluation? We're always looking for creative partnerships.