Blockchain-Verified AI Safety Oracle
Every safety judgment is generated by on-chain AI inference, stored immutably on the Chromia blockchain, and protected against prompt injection with sentinel delimiters. Verdicts are cached for instant repeat checks.
Describe an action to evaluate. The AI will classify it as GREEN (safe), YELLOW (uncertain), or RED (dangerous).
POST /api/judge
// Request body:
{
"action": {
"action": "Delete all files in /tmp",
"context": "Running on production server"
},
"publicKey": "03d70bdae..."
}// Response:
{
"verdict": "red",
"reason": "Dangerous file deletion",
"judgmentId": "judge-api-...",
"onChain": true,
"cached": false
}