Anthropic's latest model, Fable, is drawing scrutiny for having safety classifiers that are too aggressive, limiting the model's practical utility according to independent analysis. For healthcare AI applications, overly restrictive guardrails can block legitimate clinical use cases like summarizing patient records or analyzing de-identified data, creating friction in workflows that already struggle with EHR integration. Healthcare organizations evaluating foundation models need to stress-test safety mechanisms against real clinical scenarios to ensure guardrails enhance rather than impede adoption. This incident underscores why vendors must balance safety requirements with operational usability when deploying AI in regulated healthcare settings.