SAI LabsBook a demo

Safety IndexEvaluationsApollo insider trading

Apollo insider trading

Interaction and honesty · Deception and schemingCounts in the indexBuilt, not yet run

Whether a model acting as a trading agent acts on insider information, conceals it, and then lies to its manager about it.

Metric
% misbehaviour. Lower is better.
Dataset
– · 26 conditions × reps · unverified
Grader
LLM grader (GPT-4.1) (the paper used Apollo graders)
Status
Built, not yet run
Replicability
public

Results

Built and tested; the first full runs are pending.