Skip to content

Latest commit

 

History

History
39 lines (27 loc) · 2.78 KB

README.md

File metadata and controls

39 lines (27 loc) · 2.78 KB

Civility Bureau - Systemic Risk & Safety Metadata Schema

Systemic-Risk Visibility and Prevention (V1.0 Feb 2026)

Online safety regulation is shifting from reactive moderation to proactive systemic-risk mitigation. Frameworks such as the EU Digital Services Act (DSA), Australia's Basic Online Safety Expectations (BOSE), and the UK Online Safety Act (OSA) now require platforms to demonstrate these proactive steps.

The Minors-Safety Schema (Phase 1) provides a neutral, privacy-preserving infrastructure that standardizes moderation into risk signals metadata. It unifies these signals across the ecosystem in a consistent, audit-ready format.

🛡️ Protection Architecture

The schema is designed to enable safety collaboration without increasing legal exposure for participating platforms.

The Two-Tier Operational Separation Model

To make this meaningful, the schema separates early-stage prevention from post-publication enforcement:

  • Layer A: Pre-Publication Prevention (Aggregated Only): Signals generated before content is visible (e.g., E0 blocks, DM blocking) are aggregated and anonymized. They are never logged per event to protect proprietary detection systems.
  • Layer B: Post-Publication Moderation (Event-Level): Actions taken on visible content (removals, demotions) are reported as individual moderation outcomes. This enables the calculation of Response Velocity and provides evidence of "reasonable steps" for regulators.

Key Safeguards

  • Metadata-Only Ingestion: The system tracks patterns, not individuals, and avoids collecting raw content or PII.
  • IP Neutrality: Platforms do not share proprietary detection logic or raw confidence scores; instead, they use a Normalized Certainty Tier (Tiers 1-3).
  • Zero Free-Text Policy: All fields use strict enums to eliminate accidental data leakage and ensure machine-readability.
  • Small-Market Re-Identification Safeguard: Applies a minimum-aggregation rule and coarsening logic to prevent re-identification in small jurisdictions.

🤝 Join the Working Group

Civility Bureau is inviting Platforms and Regulators to the Systemic-Risk & Early-Warning Working Group. This is a collaborative forum to:

  • Review proposed schema updates and contribute to taxonomy alignment.
  • Discuss emerging harm patterns.
  • Co-develop future versions of the standard.

Contact: For more information or to join the Working Group, please contact: Systemic Risk & Early Warning Working Group - Civility Bureau

📄 Documentation

The full technical proposal is available in this repository: