{
  "title": "Responsible AI",
  "status": "published",
  "last_updated": "2026-08-31",
  "principles": [
    "assistive output is not authoritative professional advice",
    "capabilities are tenant and release gated",
    "consequential writes require policy checks and configured approvals",
    "external agents cannot self-approve",
    "unavailable or denied actions fail closed"
  ]
}