NEJM AI

Perspective by Gerald Wiest, MD, FAAN, and Oliver H. Turnbull, PhD: Faulty Artificial Intelligence, or the Sleep of Reason nejm.ai/4nV6JKh

Letter by Daniel I. Ro, MD: From Psychological Metaphors to Mechanistic Framing in Describing Errors in Large Language Models nejm.ai/4rBU9Br

Page 1 of "Response: Metaphors and Errors in Describing Large Language Models" Read the full letter at ai.nejm.org.

Gerald Wiest, MD, FAAN, and Oliver H. Turnbull, PhD, respond to a letter about their Perspective “Faulty Artificial Intelligence, or the Sleep of Reason.” Read the full response: nejm.ai/4rBUqUZ

#AI #MedSky #MLSky

Perspective by Richard K. Leuchter, MD, William B. Turner, BA, and David Ouyang, MD: Evaluating Translational AI: A Two-Way Moving Target Problem nejm.ai/3LFovDj

Letter by Richard K. Leuchter, MD, William B. Turner, BA, and David Ouyang, MD: Response to Spillover Effects in Randomized Evaluations of Translational AI nejm.ai/4cQzu8d

Page 1 of the letter "Spillover Effects in Randomized Evaluations of Translational AI" Read the full letter at ai.nejm.org.

In a letter, Sean Mann, MSc, and Carl T. Berdahl, MD, MSc, comment on the Perspective “Evaluating Translational AI: A Two-Way Moving Target Problem.” Read the full letter: nejm.ai/4seuNK1

#AI #MedSky #MLSky

Figure 1. Correctly Identified Safe and Unsafe NGT by AI in our Study (True-Positive and True-Negative Cases).

In a new study, a commercially available #AI tool for detecting nasogastric tube placement on chest radiographs demonstrated high sensitivity and negative predictive value, but its false-negative rate raised patient safety concerns. Full study results: nejm.ai/46nhZbS

#MedSky #GastroSky

Brain Health from Sleep EEG: A Multicohort, Deep Learning Biomarker for Cognition, Disease, and Mortality Sleep underpins cognition, disease prevention, and overall brain health, yet objective, integrative biomarkers of brain health remain lacking. We hypothesized that overnight sleep electroencephalog...

I’m working on rebuilding my science bubble here. Sleep/EEG/neuro/psych/biosignals folks, hello!

Starting with good news: Tried alchemy. Didn’t make gold. Made a Brain Health Score from overnight sleep EEG that tracks cognition, disease, and mortality. Philosopher’s Stone is out (NEJM AI)

Editorial Keeping humans in the loop is not a sufficient safeguard unless the right humans are empowered at the right points in the system. “Which Human-in-the-Loop? Why Context, Culture, and Health Systems Matter” by Charlotte J. Haug, M.D., Ph.D., and Ewen M. Harrison, O.B.E., F.R.C.S., F.R.S.E., F.Med.Sci.

A new editorial examines why “human-in-the-loop” oversight in medical AI is not a single safeguard but a set of roles that must be deliberately designed, and it argues that effective oversight must be adapted to local health systems, cultures, and values. nejm.ai/3OYAi0Q

#AI #MedSky #MLSky

“In the complex and uncertain world of health AI regulation, experimentation with innovative approaches is welcome. But successful experimentation requires careful evaluation of the benefits and risks of each approach and judicious policy development based on those findings.” Policy Corner “TEMPO: Experimenting with AI Sandboxes in the United States” by David Blumenthal, M.D., M.P.P.

A new article on the FDA’s TEMPO pilot to use sandboxes for selected digital devices, including health AI, cautions that while it has several appealing features, important questions about its implementation and longer-term implications remain. Learn more: nejm.ai/3OGFHJU

#AI #MedSky #MLSky

Page 1 of the Policy Corner article "TEMPO: Experimenting with AI Sandboxes in the United States" Read the full article at ai.nejm.org.

Policy Corner by David Blumenthal, MD, MPP: TEMPO: Experimenting with AI Sandboxes in the United States nejm.ai/3OGFHJU

@davidblumenthal.bsky.social #AI #MedSky #MLSky

“The [European Health Data Space] could set de facto international norms well beyond Europe, making it essential for global stakeholders to stay attuned to its evolving implementation.” Policy Corner “Driving AI Health Innovation through the European Health Data Space: Opportunities and Challenges for Non-EU Country Participation” by P. Cervera de la Cruz et al.

This Policy Corner examines the European Health Data Space, a new European Union regulation establishing a legal, technical, and ethical framework for cross-border access to and sharing of health data to accelerate #AI innovation in health care. Learn more: nejm.ai/3MEBq9o

#MedSky #MLSky

Original Article | Feb 26, 2026 Brain Health from Sleep EEG: A Multicohort, Deep Learning Biomarker for Cognition, Disease, and Mortality W. Ganglberger and Others A visual representation of the study.

This study demonstrates that an end-to-end, multitask deep learning framework applied to overnight sleep electroencephalography can derive a latent representation of brain health distilled into a single, interpretable score. Full study results: nejm.ai/4sbnrqw

#AI #MedSky #MLSky

More data isn’t magic, but it is predictable. On NEJM AI Grand Rounds, Seth Hain explains why scaling works in structured medical data. Hear more from the senior VP of R&D at Epic (@hey.epic.com): nejm.ai/ep39

#MedSky #AI #MLSky

𝗣𝗼𝗹𝗶𝗰𝘆 𝗖𝗼𝗿𝗻𝗲𝗿
Driving AI Health Innovation through the European Health Data Space: Opportunities and Challenges for Non-EU Country Participation nejm.ai/3MEBq9o

TEMPO: Experimenting with AI Sandboxes in the United States nejm.ai/3OGFHJU

𝗢𝗿𝗶𝗴𝗶𝗻𝗮𝗹 𝗔𝗿𝘁𝗶𝗰𝗹𝗲
Brain Health from Sleep EEG: A Multicohort, Deep Learning Biomarker for Cognition, Disease, and Mortality nejm.ai/4sbnrqw

External Validation of a Commercially Available AI Tool for Nasogastric Tube Position Decision Support in the NHS: A Prospective Silent Trial nejm.ai/46nhZbS

Letter: From Psychological Metaphors to Mechanistic Framing in Describing Errors in Large Language Models nejm.ai/4rBU9Br

Response: Metaphors and Errors in Describing Large Language Models nejm.ai/4rBUqUZ

𝗟𝗲𝘁𝘁𝗲𝗿𝘀
Letter: Spillover Effects in Randomized Evaluations of Translational AI nejm.ai/4seuNK1

Response to Spillover Effects in Randomized Evaluations of Translational AI nejm.ai/4cQzu8d

Cover of the March 2026 issue of NEJM AI with "NEW ISSUE NOW AVAILABLE" above it.

Volume 3, No. 3 of NEJM AI is now available! Here is a preview of the latest content:

𝗘𝗱𝗶𝘁𝗼𝗿𝗶𝗮𝗹𝘀
Which Human-in-the-Loop? Why Context, Culture, and Health Systems Matter nejm.ai/3OYAi0Q

#MedSky #AI #MLSky

Original Article by G. Starke et al.: Machine Learning–Based Patient Preference Prediction: A Proof of Concept nejm.ai/3IAIYrD

Letter by Teva D. Brender, MD, and Alexander K. Smith, MD: Machine Learning Can 𝘈𝘴𝘴𝘪𝘴𝘵 Surrogate Decision-Makers nejm.ai/3NC4IFN

Letter by Ari Nahum, MD: Model Performance Convergence Highlights Data Limitations in a Patient Preference Predictor nejm.ai/4qNgrzV

Page 1 of "Response: Machine Learning Should Enrich, Not Replace, Human Deliberation in End-of-Life Decisions" Read the full letter at ai.nejm.org.

Starke and colleagues respond to two letters about “Machine Learning–Based Patient Preference Prediction: A Proof of Concept.” Read the response: nejm.ai/45rBeQW

#AI #MedSky #MLSky

Two overlapping speech bubbles on a blue medical-themed background; one features a stethoscope. The words "emergency" and "surgery" appear in both English and Spanish.

New in NEJM Catalyst: A recent study explores the perspectives of surgical patients with limited English proficiency on AI–based and remote video interpretations, providing insights into access, equity, and barriers faced by non–English-speaking populations. Learn more: nej.md/46XDy2K

Trust isn’t assumed — it’s earned. On NEJM AI Grand Rounds, Seth Hain, senior vice president of research and development at Epic (@hey.epic.com), explains why health systems choose when and how to participate in Cosmos. Listen to the full interview: nejm.ai/ep39

#AI #MedSky #MLSky

Table 2. Chatbot Characteristics.

Table 3. Overall Performance and Individual Evaluation Criteria Scores.

Case Study by L. Uscher-Pines et al.: Assessing Generative AI Chatbots for Alcohol Misuse Support: A Longitudinal Simulation Study nejm.ai/4bi0wF2

#AI #MedSky #MLSky

“Law, ethics, and practice guidelines traditionally address those differences by prioritizing patient autonomy, clinician beneficence, fairness, and treatment utility. Yet, the patient whose health is on the line and the clinician committed to caring for them have no way of knowing what values are actually embedded in the AI systems they use.” Perspective “The Missing Dimension in Clinical AI: Making Hidden Values Visible” by C. Goldberg et al.

Perspective by C. Goldberg et al.: The Missing Dimension in Clinical AI: Making Hidden Values Visible nejm.ai/3ZvTtBe

#AI #MedSky #MLSky

Figure 1. Illustration of the Silent Validation Process.

Figure 2. Overview of the ICU Cockpit Visual Interface.

Original Article by J. Willms et al.: Silent Validation of a Longitudinal Model for Predicting Delayed Cerebral Ischemia in Real Time after Subarachnoid Hemorrhage nejm.ai/4r8KL7R

#AI #MedSky #MLSky

Page 1 of the editorial "Grading LLMs on the Ability to Grade" Read the full editorial at ai.nejm.org.

Editorial by Rebecca Sternschein, MD, MHPE: Grading LLMs on the Ability to Grade nejm.ai/49GROh1

Case Study by G. Kuling et al.: Assessment of Short-Answer Questions by ChatGPT in a Medical School Course nejm.ai/4jJFYY9

#AI #MedSky #MLSky

AI Grand Rounds Episode 39 Epic’s Approach to AI with Seth Hain A photo of Seth Hain

In the latest episode of AI Grand Rounds, Seth Hain, senior VP of R&D at Epic (@hey.epic.com), describes how his company is building foundation models that respect institutional autonomy, minimize burden, and prioritize safety. Full episode: nejm.ai/ep39

#MedSky #AI #MLSky

Figure 1. Study Flow Diagram for Human and GPT-4o Grading of Short-Answer Responses.

Figure 2. Pedagogical Grading Prompt.

Figure 3. Confusion Matrices of GPT–Human and Human–Human Grading Agreement.

Case Study by G. Kuling et al.: Assessment of Short-Answer Questions by ChatGPT in a Medical School Course nejm.ai/4jJFYY9 #ArtificialIntelligence #AIinMedicine

Editorial by Rebecca Sternschein, MD, MHPE: Grading LLMs on the Ability to Grade nejm.ai/49GROh1

#AI #MedSky #MLSky