SUI SENTINEL

Why This Matters

We are all in the age of AI. LLMs are being integrated everywhere. We can all see in the near future, AI agents will manage our finances, DeFi pools, run DAOs, and make bets on our behalf on prediction markets.

But there is one critical vulnerability: a cleverly crafted sentence can make an AI completely disregard its rules. This is called prompt injection vulnerability. As AI agents gain control over real assets and critical decisions, understanding and defending against these attacks becomes essential for the security of our digital future.

Perfect Economic Alignment

The beauty of SUI Sentinel is its economic model. It creates perfect alignment through game theory. Defenders are incentivized to deploy the most robust agents possible, because they earn SUI tokens for every failed attack attempt. The stronger their defenses, the more they earn.

Attackers are motivated by both the intellectual challenge and the potential to win the entire fund held by the agent. This creates a natural testing environment where the best security practices emerge through economic competition rather than theoretical analysis.

Verifiable Off-Chain Architecture

SUI Sentinel leverages the Nautilus framework developed by the SUI team for verifiable off-chain computation. Our AI agent server runs within a Trusted Execution Environment (TEE) using AWS Nitro Enclaves, ensuring complete isolation and tamper-proof execution.

Every AI response is cryptographically signed within the TEE and verified on-chain through our Move smart contracts. This hybrid architecture allows for complex AI computations to run off-chain for performance and cost efficiency, while maintaining blockchain-level trust and transparency through cryptographic proofs.

The result is a trustless system where players can be confident that AI decisions are authentic and unmanipulated, while enabling sophisticated gameplay that would be impossible with purely on-chain computation.

Six-Month Roadmap