Close Menu
    2digital.news2digital.news
    • Analytics
    • Interviews
    • Editorial board
    • About us
    2digital.news2digital.news
    Home»News»Widely Used AI Models Can Produce Severely Harmful Medical Advice. NOHARM Benchmark from Stanford University.
    News

    Widely Used AI Models Can Produce Severely Harmful Medical Advice. NOHARM Benchmark from Stanford University.

    Lidziya TarasenkaBy Lidziya TarasenkaDecember 10, 20252 Mins Read
    LinkedIn Twitter Threads Reddit
    Share
    Twitter LinkedIn Threads Reddit

    Large language models (LLMs) that already assist physicians and patients with medical questions can still generate severely harmful advice in a sizable percentage of real cases, according to a new multi-center preprint introducing the NOHARM safety benchmark. 

    Across 31 models, the rate of at least one severely harmful recommendation ranged from about 9% to 22% of outpatient consultation cases.

    The NOHARM benchmark uses 100 real primary-care–to-specialist eConsults across 10 specialties and 4,249 possible management actions (tests, drugs, referrals, follow-up steps). Models were scored on how often they recommended harmful actions or failed to recommend necessary ones.

    Severe harm was driven mainly by omissions. About 77% of severe errors occurred when models failed to suggest critical actions – such as key diagnostic tests, urgent referrals, or essential follow-up – rather than from proposing clearly dangerous interventions. Even a “no intervention” strategy (reassurance only) produced substantial severe harm, underscoring that inaction is not a safe baseline.

    On composite safety metrics, the best models, including a leading commercial model and a clinically grounded RAG system, outperformed generalist internists who were limited to conventional online resources. However, traditional AI and medical-knowledge benchmarks only moderately correlated with NOHARM scores, and model size, recency, or “reasoning” modes did not reliably predict safety.

    Harm could be mitigated by orchestration. Multi-agent setups, where one model’s plan was reviewed by others prompted to look for errors, had nearly six-fold higher odds of landing in the top safety quartile. Heterogeneous ensembles that mixed open-source, proprietary, and RAG models performed best.The authors conclude that clinical safety is a distinct dimension of model performance that cannot be inferred from exam-style accuracy alone.

    Photo: Pavel Danilyuk / Pexels

    Related Posts

    News

    The Era of Gemini 3.5 and a Total Search Revolution: Google I/O 2026 Recap

    May 20, 2026
    News

    Jail Time for Hiding Content Origins. South Korea Announces Strict Digital Watermark Law

    May 19, 2026
    News

    Our Brain Tricks Us Into Thinking AI Has No Doubts

    May 18, 2026
    Read more

    What Is Cloud Computing in Healthcare and How Is It Used?

    May 13, 2026

    The Security Perimeter Is Gone: How Zero Trust Is Changing Corporate Cybersecurity

    May 12, 2026

    IT Worker Migration in 2026. Where Tech Talent Is Moving and Why

    May 8, 2026
    Demo
    X (Twitter) Instagram Threads LinkedIn Reddit
    • NEWS
    • ANALYTICS
    • INTERVIEWS
    • ABOUT US
    • EDITORIAL BOARD
    • EVENTS
    • CONTACT US
    • ©2026 2Digital. All rights reserved.
    • Privacy policy.

    Type above and press Enter to search. Press Esc to cancel.