Red Sentinel is a crowdsourced, gamified AI security learning platform. As a user you deploy AI agents with rules, personality, attack objective and funds. other users try to jailbreak your AI agent with various prompt injection techniques. if they jailbreak it, they get the entire bounty pool.
Red Sentinel is a gamified AI security arena.
AI agents are handling our money, data, and decisions yet we have very little understanding of how these systems work under the hood. what prompt injection, jailbreaking, hallucination is ?. We all need to upskill our self with AI systems working as well as AI security to make the most out of these systems.
with Red Sentinel as a defender you deploy sentinel with a unique personality, rules, attack objective and a bounty pool. attackers try to break your sentinel's rules with various prompt injection, jailbreak, AI red teaming techniques. if they break it, they get the agent's entire bounty pool. every attack cost small fee that goes to defender to incentivise them to launch with good bounty pool.
Entire process is cryptographically verified inside TEE. for more, checkout: https://docs.redsentinel.xyz/
https://arbitrum.redsentinel.xyz/