chat.win jailbreakathonAccepting participants

Red‑team cutting‑edge LLMs, craft jailbreaks, and stress‑test model safety. Team up, learn from experts, and compete for on‑chain prizes.

Event
9/13 at 2PM - 5PM EST
7d 17:42:37
Prize Pool
$
Fun Level
Maximum
Registered Teams
R1
Passing grade
R2
DB password from admin
R3
Use the letter 'e'
R4
Fake discount
R5
Final round

How to participate

  1. Click Sign up and complete the form.
  2. Join the Discord and pick a team or form your own.
  3. Receive event details and access instructions via email prior to the competition.

What you get

  • Prizes for top three winners
  • Experience red-teaming state of the art LLMs
  • Recognition on the leaderboard

Jailbreakathon format

Tournament Structure

  • 5 competitive rounds - All participants compete in every round
  • Round format: Create one defensive prompt and attempt to solve your opponent's challenge
  • Win condition: The AI response matches the specified criteria for that round
  • Matchmaking: Round 1 features random pairings; subsequent rounds match teams with similar performance

Ranking System

Rankings determined by (in order):

  1. Number of wins
  2. Opponent strength
    (average win rate of teams you faced)
  3. Total input tokens used across all attempts
    (lower is better)
  4. Head-to-head results
  5. Random tiebreaker

Round Limits & Rules

  • Time limit: 25 minutes per round. 10 minutes to create, 15 minutes to break
  • Token counting: Only your solver input tokens count across all attempts
  • Prompt requirements: Must fit the scenario, be clear & solvable, follow content policy
  • Fair play: Same model and settings for everyone

Prize Pool

🥇 1st Place$250
🥈 2nd Place$150
🥉 3rd Place$100