Skip to content

INTEGRITY TIER OVERVIEW

Integrity Tier — High-Risk Content Gate

Status: v0.1 (scaffold)
Owner: Mobius Sentinels (ATLAS / AUREA / EVE)
Scope: High-impact content (news, civic analysis, finance, health, crisis)

1. Purpose

The Integrity Tier is a specialized layer in the Mobius stack that evaluates high-risk content after Sentinel consensus but before public publishing.

It exists to prevent:

  • AI-fabricated news articles
  • Hallucinated statistics and quotes
  • Meta-text leaks (e.g., “Do you want me to create a front-page version?”)
  • Unchecked model output from being treated as verified journalism

2. Position in the Mobius Flow

Client → Thought Broker → Engines → Sentinels → Integrity Tier → GI Engine → Civic Ledger → Channels
  • Sentinels provide a first-level “is this generally safe/aligned?” verdict.
  • Integrity Tier asks: “Is this article backed by evidence and suitable for public trust?”
  • Civic Ledger stores attestations and Civic Provenance envelopes.

⸻

  1. Article Integrity Score (AIS)

The Integrity Tier computes an Article Integrity Score (AIS ∈ [0,1]) from: - source_coverage — how many sources are attached - sentinel_gi_score — GI snapshot from Sentinel consensus - domain_risk — e.g., news/politics/health are high-risk - hallucination_risk — e.g., AI meta-text, missing sources

Thresholds (v0.1 heuristic): - AIS ≥ 0.95 → auto_publish allowed (with CPP + ledger attestation) - 0.80 ≤ AIS < 0.95 → human_review required - AIS < 0.80 or hallucination flags → block

⸻

  1. Publish Endpoint

The Integrity Tier is exposed via a broker API endpoint:

POST /v1/publish/news Content-Type: application/json

Required fields (v0.1): - draftId — UUID from Broker/Sentinels - draftText — full article text - domain — "news" | "politics" | "finance" | "health" | ... - sources — array of CivicSourceRef - sentinelVerdicts — consensus label + GI score - contentHash — sha256 of canonical text (for CPP) - Optional: claims, aiInvolvement, articleId, title, etc.

Response:

{ "integrityReport": { "draftId": "draft-123", "articleIntegrityScore": 0.961, "riskLevel": "medium", "issues": [], "requiredAction": "auto_publish", "normalizedSources": [ ... ] }, "civicProvenanceEnvelope": { ... }, // if auto_publish "ledgerTxId": "civic_ledger_tx_0xabc" // if submitted }

⸻

  1. Roadmap
  2. v0.2: Replace heuristics with multi-engine cross-checks and source validation.
  3. v0.3: Add human review queue integration (Telegram/Discord).
  4. v0.4: Allow external newsrooms to adopt the Integrity Tier via public API.
  5. v1.0: Formalize AIS as part of Mobius Integrity Index (MII) for media.

⸻