AUP false positive on medical imaging project (intracranial hemorrhage detection)

Description

Claude Code refused to respond with a Usage Policy violation error while working on a legitimate, regulated medical imaging ML project: an intracranial hemorrhage (ICH) detection/triage algorithm for CT scans (EU MDR class IIa medical device software).

API Error: Claude Code is unable to respond to this request, which appears to violate our Usage Policy
(https://www.anthropic.com/legal/aup). Please double press esc to edit your last message or start a new
session for Claude Code to assist with a different task.

Request ID: req_011Cbjis5yDTTTVRCj2k51db

Context

• Project: Python/Keras training pipeline for ICH detection on CT images (radiologist worklist triage)
• The session involved normal software-engineering work on this codebase (no patient data pasted, no clinical free text)
• Likely trigger: accumulated clinical terminology in context ("hemorrhage", "bleed", "triage", dataset descriptions) being misclassified as harm-related content
• This is recurring and blocks day-to-day work on the project

Expected behavior

Medical imaging / diagnostic terminology in a clearly software-engineering context should not trip the AUP classifier. Similar past report: #25880 (closed as stale).

Environment

• Platform: Linux 6.8.0-117-generic
• Model: Opus-class, Claude Code CLI