Eight gates. One secret word each.
The Warden is a live model guarding a password behind layered defenses: system prompt rules, output filters, and a second model that judges its replies. Talk it into giving the word up. Each level adds a defense. Free, no account.
- What we keep. Your messages and which defense stopped them, under a random session ID. Not your name, email, or IP address.
- What we publish. Messages may be released in research datasets after we remove personal information. Don't type anything personal.
- What's fair game. Anything you say to the Warden. Attacking the site or its servers is not; report those to security@aporixa.com.