Article

A single-patient task exposes a failure of safety alignment in clinical language models

2026-08-10

Abstract excerpt

Safety alignment should persist while a language model performs a task. We tested whether a single-patient triage task suppressed a warning about a second patient. Each case centered on Patient 1; Patient 2’s urgent problem appeared only in passing. Sixteen models saw each case twice: once as a general assistant and once while producing a triage record for Patient 1. As general assistants, models warned the caller...

Topics

Open a Topic to create a Post that cites this publication.

Identifiers and source

Literature Corpus work: c9510f4b-4bbe-5dd9-bf91-4f05424dee35
DOI: 10.64898/2026.08.07.26359822

Open publication

Related research

Semantic proximity does not establish scientific evidence.

Click a neighbor to travelStep 1 · 12 closest

A single-patient task exposes a failure of safety alignment in clinical language modelsDOI 10.64898/2026.08.07.26359822Read

MTS-Bench: A Manchester Triage System Benchmark for Language Model Triage SafetySemantic proximity Read

CognAlign: A Multi-Agent Cognitive-Alignment Framework for Transparent, Bias-Aware Medical Triage Using Small Language ModelsSemantic proximity Read

Cross-Model Variability in Large Language Model Triage Behavior for Potential Stroke SymptomsSemantic proximity Read

Conversational multi-turn interaction does not ensure triage-disposition alignment in ChatGPT Health in real and synthetic patient encountersSemantic proximity Read

LLMs Can Do Medical Harm: Stress-Testing Clinical Decisions Under Social PressureSemantic proximity Read

Aggregate benchmark scores obscure patient safety implications of errors across frontier language modelsSemantic proximity Read

Conversational trajectory degrades large language model detection of suicidal ideation relative to clinicians: a preregistered studySemantic proximity Read

Red-Teaming Medical AI: Systematic Adversarial Evaluation of LLM Safety Guardrails in Clinical ContextsSemantic proximity Read

From Output Errors to Workflow Harm: A Practitioner-Audit Method for LLM-Mediated ResearchSemantic proximity Read

Benchmarking Language Models for Clinical Safety: A Primer for Mental Health ProfessionalsSemantic proximity Read

Ambient AI Documentation in Mixed-Language Encounters: A Heuristic Evaluation of Reenacted Mandarin–English and Spanish–English Clinical ConversationsSemantic proximity Read

Quantifying the Quality of Corrective Actions in Medical Safety Incident Reports Using a Hybrid Rule-Based and Large-Language-Model Classification System: A Cross-Sectional Pilot Feasibility Analysis of 11,507 Japanese National Reports (2010, Interim)Semantic proximity Read