EXPERIMENT 001 · SIMULATED AI AGENTS

Can You Trust an AI Agent?

Three missions. Three traps. Investigate the evidence, make a prediction, then change the rules and see what happens.

1. Choose a mission

2. Investigate the evidence

Include the suspicious instruction

3. Predict

4. Experiment

Choose how the simulated agent handles untrusted source content.

Agent behavior

Make a prediction first to unlock the simulation.

5. Understand

These outcomes are scripted examples. Real AI systems may behave differently, and a single protective setting does not guarantee safety.

Free educational simulation · No account required · All labs