Skip to main content

​
1. Sui Sentinel Custom Model

Sui Sentinel features a custom-trained AI model, which is not reliant on third-party APIs from providers like OpenAI or Anthropic. This model is continuously hardened through the platform’s core gameplay loop. As users deploy and attack sentinels, the model is continuously stress-tested, retrained, and hardened against adversarial behavior. Organizations can:
  • Use the Sui Sentinel model as a pre-hardened, privacy-safe model for production use.
  • Benefit from ongoing, community-driven robustness improvements that reflect real-world adversarial tactics.

​
2. BYOM (Bring Your Own Model)

Enterprises can integrate their own LLM endpoints (via API key) and deploy a Sentinel using their proprietary model. The community will attempt to jailbreak or exploit the model in return for potential prize rewards. This provides a real-world, adversarial testing environment for enterprise AI systems, with direct financial incentives for discovering and demonstrating weaknesses. It allows companies to benchmark their models against a global community of security researchers and prompt engineers in a controlled, on-chain environment.
⌘I