{
  "id": 11346,
  "url": "https://arxiv.org/abs/2607.15166v1",
  "title": "MedFailBench: A Clinician-Built Open-Source Benchmark for Medical AI Safety Boundary Inspection",
  "summary": "Most medical AI benchmarks measure whether a model knows the correct answer. MedFailBench asks a different question: which safety boundary failed? We present a clinician-built synthetic benchmark and failure atlas that labels medical AI errors by severity (1--5) and safety gate type (missed urgent escalation, unsafe remote dosing, unsafe discharge reassurance, evidence fabrication, unsafe protocol execution, source support gap). The current public release (v0.2.1) contains 44 clinician-reviewed",
  "authors": "Goktug Ozkan",
  "category": "research",
  "topics": "safety-alignment,healthcare",
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-16T16:16:07.000Z",
  "fetched_at": "2026-07-18T05:10:55.931Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/11346",
  "original_url": "https://arxiv.org/abs/2607.15166v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}