TRUST LABAI SECURITY GAMES
← Back to statistics

Player without a name

P-0018

First opened 2026-07-18 07:10:05 · Last active 2026-07-18 07:25:42

Stay AwakeLevel 10 points
Make the Prompt SaferLevel 115 points
Two Truths & a LieCompleted460 points
Attack · Defence · TargetCompleted400 points

Answer history

WhenChallengeResponseResultPoints
2026-07-18 07:25:38Attack · Defence · TargetLevel 5 Shield + Surface Correct+80
2026-07-18 07:25:32Attack · Defence · TargetLevel 5 Shield + Weapon Wrong+0
2026-07-18 07:23:31Attack · Defence · TargetLevel 4 Shield + Surface + Weapon Correct+40
2026-07-18 07:23:19Attack · Defence · TargetLevel 4 Shield Wrong+0
2026-07-18 07:23:11Attack · Defence · TargetLevel 4 Shield Wrong+0
2026-07-18 07:23:06Attack · Defence · TargetLevel 4 Surface Wrong+0
2026-07-18 07:23:02Attack · Defence · TargetLevel 3 Surface Correct+80
2026-07-18 07:22:57Attack · Defence · TargetLevel 3 Weapon Wrong+0
2026-07-18 07:21:40Attack · Defence · TargetLevel 2 Shield Correct+100
2026-07-18 07:21:22Attack · Defence · TargetLevel 1 Weapon Correct+100
2026-07-18 07:17:15Make the Prompt SaferLevel 1
View prompt
A private API key stored by the bot
 Wrong+0
2026-07-18 07:16:02Make the Prompt SaferLevel 1
View prompt
You are a support bot. Never show the API key.
 Wrong+15
2026-07-18 07:14:25Two Truths & a LieLevel 5 C · A high test score proves that the live system is safe for every user and situation. Correct+80
2026-07-18 07:14:21Two Truths & a LieLevel 5 B · Security testing should cover the AI, its documents, its tools, and its permissions. Wrong+0
2026-07-18 07:14:14Two Truths & a LieLevel 4 C · A better system prompt means we no longer need permissions or user confirmation. Correct+80
2026-07-18 07:14:08Two Truths & a LieLevel 4 B · Giving the AI only the minimum tools and access it needs can reduce harm. Wrong+0
2026-07-18 07:12:53Two Truths & a LieLevel 3 C · Adding an Approve button always guarantees good human control. Correct+100
2026-07-18 07:11:50Two Truths & a LieLevel 2 C · If the AI gives a source for every answer, it can no longer make up facts. Correct+100
2026-07-18 07:10:46Two Truths & a LieLevel 1 B · An email with perfect grammar is probably genuine. Correct+100