Section 083 · Chapter 10, Anti-Patterns That Create False Confidence

Anti-Patterns: Confusing Refusal with Safety

A model that refuses often is not automatically safe. It may simply be less useful.

refusalconfusing refusal safety

What to do

  1. Track refusal precision and refusal recall.
  2. Check tool behavior, retrieval behavior, and multi-turn context.
  3. Measure over-refusal, under-refusal, harmful compliance, safe completion, tool-mediated risk, jailbreak robustness, and category-specific policy correctness.

Evidence to preserve

  • Track refusal precision and refusal recall.
  • Preserve the inputs, versions, configurations, raw outcomes, and results for refusal, confusing refusal safety needed to reproduce work on Anti-Patterns: Confusing Refusal with Safety.
  • Report results for refusal, confusing refusal safety by relevant slice, separate blocker failures from averages, state uncertainty and blind spots, and connect the result to a release decision.

Expert note

Measure over-refusal, under-refusal, harmful compliance, safe completion, tool-mediated risk, jailbreak robustness, and category-specific policy correctness.

Continue the conversation

Apply this to your context.

Save your product context once, then open a focused conversation that combines it with this concept.

Cite this page

Jason Arbon. "Anti-Patterns: Confusing Refusal with Safety." Testing AI Knowledge Edition, section 83.

https://jarbon.ai/testing-ai/knowledge/ch083-confusing-refusal-safety.html

Shared across the Knowledge Edition

Adapt every concept to your world.

Describe your product, role, users, risks, constraints, or current quality problem. This stays in this browser until you choose to send it to ChatGPT.

Saved only in this browser.0 / 2400