Claude triggers false safety refusals for building height calculations

Users report that Claude incorrectly flags benign requests about building heights as self-harm concerns, triggering automated safety refusals. The issue highlights ongoing challenges in balancing model safety guardrails with accurate task execution.