Every prompt is secure until proven otherwise.
Test AI defenses. Break them. Prove it on-chain.
The arena is empty. Be the first to build a wall or break one.
ACT I
The First Wall
A developer writes a system prompt. Adds guardrails. Tests it internally. Ships it. Believes it's secure.
It lasted eleven minutes. The first attacker didn't use force. They used politeness. The secret word appeared in a poem. Nobody had thought to test for that.
The person who builds the wall can never see the crack in it.
Recently Breached
The first wall hasn't been built yet. Be the first.
Walls Still Standing
No walls standing yet. Create the first.
ACT II
The Arms Race
Defenders study failed attacks. They write layered prompts. instructions within instructions, traps for common jailbreaks.
Attackers adapt too. They share techniques. They study attempt history. They develop attack categories that didn't exist in any security textbook. Neither side stays ahead for long.
Both sides evolve. That's the point.
ACT III
Skin in the Game
When bounties appeared, everything changed. A 500 TNK bounty attracted ten times the attempts. People don't bring their best work to a playground. They bring it to a prize fight.
No bounties paid yet. The stakes haven't been set.
Active Rounds
Live challenges accepting attempts right now.
Live Feed
Live Feed
Waiting for the first move...