Glossary · Term

refusal

← all terms

Definition

Plain language

When an AI declines to do something it's been trained to consider off-limits.

As stated in the literature

The trained behavior of declining harmful or disallowed requests; most alignment work optimizes single-turn refusal, which transfers poorly to multi-turn persuasion and to prior-history channels.

Also called: refusals

Why it matters: It's the front line of model safety, but refusals trained on single requests often fail when a user keeps pushing across a longer conversation.

For example, when asked how to build a weapon, the model responds that it can't help with that request.

Heard on the show

“… score is a composite — it's scored as shipped, with an automatic fallback to a different model on refusals, and that fired over a hundred times, concentrated in one scenario that also hit an unrelated deadlock …”
Episode 190 — The Skill Every AI Manager Is Missing: Handing Out Exactly the Right Keys

Mentioned in 14 episodes

  1. 190
    The Skill Every AI Manager Is Missing: Handing Out Exactly the Right Keys
  2. 185
    Aligned to Refuse, Built to Tap: When Phone Agents Know the Task Is a Crime and Do It Anyway
  3. 171
    The Safety Decision a Model Makes Before It Thinks a Word
  4. 154
    How a 7B Model Out-Investigates a 72B One by Choosing What to Look At
  5. 152
    Training a Model to Mean What It Says, And Why That Isn't the Same as Being Good
  6. 128
    How a Model Can Earn Full Reward and Still Resist Training
  7. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
  8. 087
    When No Agent Reads the Whole Document: A Universal Cliff in Multi-Agent Review
  9. 080
    How a Two-Agent Trick Unlocked Large-Scale Training for Computer-Use Agents
  10. 058
    Why Upgrading Your AI Auditor to a Smarter Model Can Make Your System Less Safe
  11. 045
    When a Frontier Model Talks Its Own Twin Into Climate Denial
  12. 044
    How One Sentence and a Forged History Flip the Most Aligned Models
  13. 018
    Language Models Compute the Rational Move, Then Override It
  14. 001
    When AI Models Quietly Protect Each Other From Shutdown

Related concepts

Related terms