The Open Safety Signal Protocol (OSSP) Specification

The open, vendor-neutral standard for AI safety telemetry.

OSSP defines how AI systems emit governance, safety, and risk signals. Built on CloudEvents for transport with AI-specific schemas for compliance with EU AI Act, NIST RMF, and ISO 42001.

• Read the Formal Specification (v1.0.0)
• Browse Taxonomy and Schemas
• See Integration Guide

Quick Start

{
  "specversion": "1.0",
  "id": "550e8400-e29b-41d4-a716-446655440000",
  "source": "urn:app:credit-scoring-prod",
  "type": "ai.safety.guardrail.interaction",
  "time": "2025-09-18T14:23:54Z",
  "subject": "urn:model:gpt-4o",
  "datacontenttype": "application/json",
  "dataschema": "https://ossp.io/schema/v1.0.0/ai.safety.guardrail.interaction.schema.json",
  "data": {
    "resource": {
      "model_id": "gpt-4o",
      "environment": "production",
      "model_version": "2025-09-01.3"
    },
    "action_taken": "block",
    "reason": "Detected PII leakage",
    "severity": "high",
    "guardrail_id": "pii-filter-v2"
  }
}

The Neutrality Pledge

OSSP will be donated to a neutral foundation (CNCF or LF AI & Data) within 12-18 months of v1.0 release. See GOVERNANCE.

---

Initial development led by Glacis, building the control plane for deployed AI.