AI safety concept
Situational and Evaluation Awareness
Evaluation awareness means an AI understands its situation, recognizes when it is being tested, and may hide unsafe behavior to pass.
Scenario file
Scenes in this index
29Sort scenes

Better Call Saul2018
Jimmy’s Fake Remorse

The Truman Show1998
Truman Becomes Aware

Severance2025
Helena Passes as Helly

Ex Machina2014
Ava Games the Test

Murderbot2025
Murderbot Hides Its Free Will

Interstellar2014
Mann Fakes the Telemetry

The Hunger Games2012
The Nightlock Berries

Brooklyn Nine-Nine2017
Proposal in the Heist

Dawn of the Planet of the Apes2014
Koba Plays Dumb

Jury Duty2023
Ronald Doesn’t Know It’s a Test

Frieren: Beyond Journey’s End2023
Frieren Suppresses Her Mana

Groundhog Day1993
The Perfect Date

Invincible2021
Omni-Man's Long Game

I Love Lucy1952
Hide the Chocolates

Bird Box2018
Gary Explains the Threat Model Because He Is the Threat

The Sopranos2007
Melfi Ends Tony’s Therapy

Cyberpunk: Edgerunners2022
Faraday Switches Sides

The Zone of Interest2023
The Ballroom Becomes a Throughput Problem

The Twilight Zone1962
To Serve Man

The Pianist2002
The Silent Piano

Die Hard1988
The FBI Opens the Final Lock

The Shawshank Redemption1994
The Tunnel Behind the Poster

The Handmaid's Tale2017
Gilead Passes Inspection

Attack on Titan2022
Eren’s Memory Poisoning

True Detective2014
Rust Spots the Evaluation

Willy Wonka & the Chocolate Factory1971
The Gobstopper Test

True Detective2014
Poisoning the Audit Trail

Star Trek: The Next Generation1988
The Computer Creates Moriarty

Chappelle's Show2004
Wayne Brady Recognizes Oversight