Glossary · Term

false positive

← all terms

Definition

Plain language

A false alarm — flagging something as a problem when it actually isn't.

As stated in the literature

An instance incorrectly classified as positive by a detector; the rate of these trades off against recall and is a key cost metric for safety monitors and guards.

Also called: false positives, false-positive

Why it matters: Too many false alarms make a safety tool annoying and untrustworthy, so keeping them low is a key cost of any detector.

For example, a spam filter that moves an important work email to the junk folder has produced a false positive.

Heard on the show

“How many wrong answers are lying around, and how often the judge waves a wrong one through — call that second one the false-positive rate.”
Episode 207 — An AI Graded Its Own Math Test 94 Percent — It Actually Scored 20

Mentioned in 21 episodes

  1. 207
    An AI Graded Its Own Math Test 94 Percent — It Actually Scored 20
  2. 202
    How Do You Know an AI Agent Actually Refused? Check the World, Not the Words
  3. 201
    One in Four NeurIPS Papers Cites a Reference That Doesn't Exist
  4. 181
    How to Backpropagate Blame Through a Team of Chatbots — And When It Backfires
  5. 159
    Can a Coding Agent Run Its Own Robot Experiments Overnight, With No Human Resetting the Scene?
  6. 143
    When a Model Notices You Forged Its Own Words, And Why That Breaks Safety Tests
  7. 133
    How MiniMax Turned a Reward-Hacking Disaster Into Olympiad Gold
  8. 128
    How a Model Can Earn Full Reward and Still Resist Training
  9. 124
    A Cheap Model With the Blueprints Beats Expensive Models Working Blind
  10. 118
    Why the Best-Aligned AI Models Are the Easiest to Trick Into Producing Harm
  11. 111
    How a 4B Web Agent Beat Models 60x Its Size on 500 Demonstrations
  12. 105
    The Trojan Is Your Agent's Memory: Why Single-Step Defenses Miss Persistent Attacks
  13. 102
    How to Catch an AI Attack That No Single Conversation Reveals
  14. 089
    When AI-Written Papers Read Well But the Evidence Underneath Is Broken
  15. 087
    When No Agent Reads the Whole Document: A Universal Cliff in Multi-Agent Review
  16. 062
    Treating Hallucinations as Exploits: A Gate-Based Architecture for Agent Safety
  17. 057
    How Uber Caught 206 Leaked Credentials With an LLM-Powered Security Stack
  18. 039
    When Smarter Agents Get Fooled by Three Extra Nodes in a Database
  19. 024
    An AI Agent That Found 28 Zero-Days in Windows — And What Made It Work
  20. 014
    Why a Constrained Pipeline Beat a Full Coding Agent at Finding Bugs 30-to-1
  21. 004
    The Sycophancy Circuit That Survives Alignment Training

Related concepts

Related terms