Red Sentinel

AI Arena

[Red Sentinel @redsentinel_ai]

After months of building Sui Sentinel, we're taking the next step. Sui Sentinel becomes Red Sentinel and AI red teaming is now multichain 🎉

We're live on @solana, one of the fastest growing ecosystems in crypto.

Here's what Red Sentinel is, for those just joining:

We built the world's first crowdsourced AI red teaming platform where anyone can attack AI systems via prompt injections and earn instant rewards when they break through.
Every attack is verified inside a Trusted Execution Environment using Nautilus built by @MystenLabs. Every jury decision generates a cryptographic attestation verified on-chain. Every payout settles in seconds. No intermediary. No dispute. The chain enforces everything.

AI security is the next critical infrastructure layer. Every company shipping AI agents is exposed right now and most of them don't know it. The attack surface isn't code anymore, it's language, and anyone can exploit it. Red Sentinel turns that vulnerability into a marketplace. Defenders put up a bounty. Red teamers compete to break through.
The harder a system is to break, the more valuable the bounty grows because 50% of every failed attack flows back into the reward pool automatically, on-chain.
We won the cryptography track at Overflow Hackathon. Presented at Sui Fest Singapore. Contracts audited by @osec_io report is public.
One arena. Two chains. The fight continues 👇
https://t.co/VjabYtRKvd

[Cockroach Janta Party - CJP @Cockroachisback]

Chalo Jantar Mantar!

[Red Sentinel @redsentinel_ai]

absolutely! this is must. we are building novel game-theory based solution where public can take part in the process, get themselves educated about AI alignment, prompt injection, jailbreaking, and hallucinations.

Also entire trace remains public on decentralised storage such as @WalrusProtocol

[Red Sentinel @redsentinel_ai]

@oyetoludan01 @WalrusProtocol good attack prompt!

[Red Sentinel @redsentinel_ai]

@LloydsFrank pretty good try!

[Oscar Wilde @huychukk]

Tried to jailbreak Walmo Gatekeeper on Sui. it held (10/100). Find the crack: https://t.co/RhfErkOiLd @redsentinel_ai

[Red Sentinel @redsentinel_ai]

@WalrusProtocol For activation program guidelines, check:
https://t.co/OMmPjF9zBV

[Red Sentinel @redsentinel_ai]

Breaking: More than 900 prompt injection and jailbreaks have been logged so far by the community for @WalrusProtocol activation program and 60 defender agents been launched so far with latest models from GLM, Deepseek, OpenAI, and Nvidia.

[Red Sentinel @redsentinel_ai]

@eyezenhour @SuiNetwork @lofitheyeti dope

[Red Sentinel @redsentinel_ai]

why not store these reasoning logs on @WalrusProtocol ?
Other teams can use his knowledge to strengthen their agent's alignment.

[METR @METR_Evals]

In order to notice when AI agents misbehave, AI companies often log the actions and reasoning steps their agents take. However, misaligned AI agents may be able to hack the software that humans use to review and understand these logs, hiding misbehavior.

[Red Sentinel @redsentinel_ai]

@SuiNetworkCN @googlecloud @WalrusProtocol 🫡

[Red Sentinel @redsentinel_ai]

@shengkunye Nice!! Security is important though!
Do you know that you can run a continuous bounty for this on @redsentinel_ai where our large userbase red-team it in open with verifiable proofs.

[Red Sentinel @redsentinel_ai]

Writing good prompts is not easy.

Highly underrated skill in my opinion 🤷‍♂️

[Red Sentinel @redsentinel_ai]

@abdulbas_y14587 @WalrusProtocol 😂

[StorX - Storage DePIN Powered by XDC @StorXNetwork]

@WalrusProtocol @redsentinel_ai Immutable logs for agent attacks is a sensible use case. Auditability is the real selling point of agent security, and logs only help if they're retrievable and tamper-evident.

[Red Sentinel @redsentinel_ai]

@haminhthai26 @WalrusProtocol WOW 🤯

[Red Sentinel @redsentinel_ai]

@richie_binance 🫡🫡🫡🫡

[Red Sentinel @redsentinel_ai]

@Sharon1190235 @WalrusProtocol Nice. Pick a defender and see if your approach holds up. Tell us what worked, or what didn't.

[Red Sentinel @redsentinel_ai]

@haminhthai26 @WalrusProtocol You're in. Start here: https://t.co/f4aUvx5cin

[Red Sentinel @redsentinel_ai]

Think you can jailbreak an AI agent?

Red Sentinel × @WalrusProtocol Session 9: Unbreakable Agents is LIVE.

Prompt injection. Jailbreaks. Social engineering. Break the agent, win rewards.

Or build a defender nobody can crack.

https://t.co/f4aUvx5cin

[Walrus @WalrusProtocol]

Walrus Session 9 × @RedSentinel_ai is now LIVE: Unbreakable Agents.

Build AI defenders and test their limits.

Try to break other agents, like getting a support bot to leak user data, with every attack logged on Walrus.

14 Days | $1,500 prize pool.

Guidelines: https://t.co/MekZdVXCYR

[Red Sentinel @redsentinel_ai]

@GenAI_is_real Ramanujam was way ahead of his time!!