Trust Intelligence Infrastructure

AI produces intelligence. Censio produces trust.

Censio issues the verifiable record of what your AI concluded — calibrated confidence, complete evidence, independent proof — so enterprises can act on AI, and prove it years later.

Read the API

Developer, enterprise & individual access — waitlist open

THE EVIDENCE NETWORK · TIER 1 PUBLIC · TIER 2 COMMERCIAL · TIER 3 ENTERPRISEEXPLORE THE DATA →
BloombergMARKETS & FINANCIAL DATA
ReutersNEWS & EVENTS
S&P GlobalRATINGS & INDICES
Associated PressNEWS WIRE
NewsGuardSOURCE CREDIBILITY
PolymarketPREDICTION MARKETS
Bright DataWEB & SOCIAL DATA
FREDMACROECONOMIC

MARKS INDICATE TARGET INTEGRATIONS ACROSS A MAPPED ~100-PROVIDER ECOSYSTEM — CAPABILITY, NOT PARTNERSHIP OR ENDORSEMENT · SUBJECT TO LICENSING

01 · What Censio is

The system of record for machine judgment.

Censio issues the record of what an AI concluded — the evidence, the method, the moment — verifiable by anyone, forever. One calibrated confidence score on the surface; the full trust profile, version history and independent proof underneath.

Works above · OpenAI · Anthropic · Google · Meta · Mistral · any model

02 · The stack

Between AI and the decisions made on it.

Censio is not another model. It is the layer that sits between AI systems and enterprise decisions — measuring trust, issuing the record, proving it forever.

SYSTEM DIAGRAM · CEN-ARCH-01 AI MODELS OpenAI · Anthropic · Gemini Llama · Mistral · any THE EVIDENCE NETWORK TIER 1 · PUBLIC & OPEN TIER 2 · COMMERCIAL TIER 3 · ENTERPRISE PRIVATE Real-time · ~100 mapped providers ONE API · ONE MCP CENSIO TRUST INTELLIGENCE ENGINE Evidence weighting · source reliability Contradiction detection Calibrated confidence · explainability MEASURES · NEVER RECOMMENDS TRUST OBJECT The governed, machine-readable record of machine judgment CX-924837 · SHA-256 SEALED DECISIONS & AGENTS Your policy · your thresholds AI agents natively via MCP INDEPENDENT VERIFICATION · TIP Dual-rail anchoring · chain-agnostic · verify without trusting the issuer — censio.io/verify
AI MODELSOpenAI · Anthropic · Gemini · any
EVIDENCE NETWORKTier 1 public · Tier 2 commercial · Tier 3 enterprise
CENSIO ENGINEOne API · one MCP — confidence · contradiction · explainability
TRUST OBJECTThe governed, machine-readable record
DECISIONS & AGENTSYour policy · your thresholds · native via MCP
VERIFICATIONIndependent · dual-rail · never via the issuer

Every AI decision deserves measurable confidence, explainable reasoning and independently verifiable integrity.

03 · The platform

Four capabilities. One primitive.

01
Trust IntelligenceThe engine

Evidence collection and weighting, source reliability, contradiction detection, calibrated confidence — every score explainable, every methodology versioned and archived.

02
Trust ObjectsThe asset

Every assessment becomes one governed record: claim, confidence, evidence, methodology, timestamp, version history. Machine-readable trust; Censio Certs are its branded rendering.

03
VerificationThe proof

Independent integrity infrastructure confirms every record existed, unaltered, at a moment in time — without trusting Censio’s word for it. CensioCert™ blockchain anchoring: active partner discussions underway.

04
GovernanceYour policy

Act above your threshold, escalate below it, require review on contradiction. Censio measures — you decide. The record proves both.

04 · The primitive

Not a certificate. A Trust Object.

A certificate is a static document. Trust is dynamic. Every Censio evaluation issues a structured, living record —

  • one calibrated confidence score, full trust profile inside
  • the complete evidence set — support, conflict, provenance
  • methodology version, model panel, timestamp, issuer
  • version history preserved on an unbroken chain
  • independently verifiable by anyone, forever

The CensioCert™ is the certificate layer of a Trust Object — the document your auditor holds and your counterparty trusts. The API is its machine form; the public verify page its public face. One record, three surfaces.

TRUST OBJECTCX-924837
AssessmentSupplier compliance review
Confidence94%
Evidence sources52
ContradictionLow
Methodologyv3.1
Version12 · history preserved

05 · CensioCert™

The certificate layer. CensioCert™.

Every Trust Object can be issued as a CensioCert™ — the certificate your auditor holds and your counterparty trusts. Grounded in the live evidence network, sealed under TIP, fingerprint-anchored on-chain — and checkable by anyone holding the QR, forever. Each certificate’s fingerprint becomes an immutable, append-only hash record — auditable end to end; blockchain-anchoring partner discussions are ongoing.

For anyone — no account, no trust required
  • Work beside AI all day? Check the answer before you act on it — and attach the CensioCert™ when you need to show your work
  • Scan the QR on any CensioCert™ — see the claim, the confidence score and the evidence behind it
  • The check never routes through Censio’s servers (TIP-4) — you are not taking our word for it
  • If the record has been superseded or revoked, the certificate says so instantly
For enterprises — at scale, under your policy
  • Automated decision-making — gate actions on /policy/check before they execute; every decision leaves a certificate behind
  • Data integrity — monitored supersession, unbroken version chains, audit-pack export for the regulator
  • Blockchain issuance — epoch-anchored fingerprints on public rails; the evidence never leaves your control
Automated decision-makingData integrityCompliance evidenceCounterparty diligenceRegulator replayVerify one now →

CensioCert™ · issued under TIP · grounded in the evidence network · specimen live at censio.io/verify

06 · The protocol

Codified as TIP — the Trust Intelligence Protocol™.

The full life-cycle of a machine-judgment record, published as an open five-layer specification. TIP is Censio’s invention — published as an open specification, with Censio as the first issuer and reference implementation. Specifications can be copied; the issuance history cannot.

TIP-1
RECORD
The Trust Object schema — claim, calibrated confidence, evidence set, methodology version, model panel, version chain.
TIP-2
SEAL
Canonical JSON → SHA-256 fingerprint. One record, one fingerprint — deterministic from any conforming implementation.
TIP-3
ANCHOR
Epoch batching: fingerprints → Merkle tree → root anchored dual-rail — an eIDAS-qualified timestamp plus a public-chain transaction. Only the root touches the chain; never the evidence, never customer data.
TIP-4
VERIFY
Recompute the fingerprint, walk the Merkle inclusion proof to the anchored root, check the chain transaction and the qualified timestamp. Never routes through the issuer.
TIP-5
SUPERSEDE
Living records: version chains, supersession notices, revocation — trust is dynamic, and the protocol carries the history.

Around the constitution run six extension protocols — all Censio inventions, all already operating inside the platform: temporal trust, decision replay, machine warrants, outcome witnessing, evidence custody and cross-issuer federation. Codified as the TIP extension family; full specifications publish progressively.

Published as an open specification · MCP tools map 1:1 onto TIP-4 / TIP-5 · read the full specification → · see it live →

07 · API

Claims in. Trust Objects out.

Censio is programmable from day one. One integration, and every AI response your systems rely on becomes trust-aware.

POST/evaluate
GET/trust-object/:id
POST/verify
GET/history/:id
POST/policy/check
POST/monitor
// every response becomes trust-aware
{
  "answer": "Approved under your policy",
  "confidence": 91,
  "trust_object": "CX-829384",
  "verification_status": "verified",
  "methodology": "v3.1",
  "evidence_count": 37,
  "contradictions": 1,
  "version": 12
}

08 · MCP

Agents consume trust natively.

AI is becoming agent-based. The Censio MCP server lets any agent — Claude, Cursor, GPT, anything that speaks MCP — verify claims mid-workflow and act against your thresholds.

"mcpServers": {
  "censio": { "url": "https://mcp.censio.io" }
}
TOOLverify_claim — score a claim in-flow
TOOLget_trust_object — fetch the record
TOOLcheck_policy — meets your threshold?
TOOLmonitor_claim — subscribe to change

09 · In practice

Where the record earns its keep.

Financial services · compliance

The supplier review that survives the regulator.

A bank’s AI reviews a supplier. Censio issues the Trust Object — 94% confidence, 52 sources, methodology v3.1, sealed and anchored. Three years later the regulator asks what supported the decision: the bank replays the exact record and proves it independently.

Insurance · claims governance

Every claims decision carries its evidence.

Claims AI acts only above the insurer’s own confidence threshold; below it, the record routes to review. Contradiction in the evidence flags automatically. The Audit Pack exports objects, proofs and archived methodology in one regulator-ready bundle.

AI agents · MCP

Agents verify before they act.

Mid-workflow, an agent calls verify_claim on a market-moving assertion and receives calibrated confidence with cited evidence — not a guess. When the record is superseded, the agent is notified and re-checks.

Procurement & legal · diligence

Counterparty claims, checked against the registries.

Ownership, filings, sanctions exposure and litigation history assessed across Tier-1 registers and commercial intelligence — issued as Trust Objects that travel with the contract and stay verifiable for its life.

10 · Autonomy

Trust at machine speed.

Evaluation is deliberate. Enforcement is instant. Three primitives take Censio into automated, high-stakes environments — trading systems, claims engines, critical infrastructure — where decisions happen faster than any human review.

Censio Mandates · the time-based decision layer

Time-boxed authority to act.

A Mandate binds a Trust Object to your policy with a validity window and a cryptographic signature. Systems verify it locally in microseconds — no network on the hot path. When the window lapses or the evidence supersedes, the mandate dies on its own.

Protocol: MWCP — Machine Warrant Chain Protocol™

Sentinel SDK · shadow mode

Install it. Change nothing. It watches.

The Sentinel SDK sits beside your application as a shadow layer: every AI-assisted decision quietly leaves a record from day one. When you’re ready, flip to mandate mode and the same layer enforces your thresholds before actions execute.

Protocols: TIP · MWCP™ enforce · CDRP™ records

The Black Box · decision replay

AI is a black box. Censio is the black-box recorder.

Every decision, its mandate, its record and the evidence as it stood — anchored, frame by frame. Scrub back to any moment, prove what the system knew when it acted, and re-run it under the archived methodology. Replay isn’t a simulation; every frame is provable.

Protocol: CDRP — Chronological Decision Replay Protocol™

Financial marketsMarket-making desksClaims automationCritical infrastructureConnected & autonomous vehiclesDefence-grade environmentsMandates — drafted as the TIP enforcement profile

11 · Enterprise

Prove what your AI concluded. Years later.

A regulator asks what analysis supported a decision made three years ago. "It's in our database" is not an answer — it asks them to trust you. Censio's answer: here is independent proof this exact record existed, unaltered, at that moment.

Built for financial services and insurance first — where auditability of AI-assisted decisions is becoming procurement's first question.

  • audit-ready records for every AI-assisted decision
  • your policies, your thresholds — enforced and evidenced
  • litigation defence: prove nothing changed after the fact
  • verification independent of Censio's infrastructure
  • EU AI Act record-keeping, by construction

12 · The evidence network

Real-time enterprise data. One API. One MCP.

The Trust Intelligence engine is only as strong as the evidence it analyses — so Censio does not simply consume data, it builds an ecosystem of trusted providers, streamed through one endpoint. Every connected source strengthens confidence computation, contradiction detection and the evidence graph behind every Trust Object. The objective is not the most data. It is the most trusted data.

Tier 1 · Public & open

Government registers, regulators, academic research, standards bodies, open knowledge, public news. The foundation of the evidence graph.

Tier 2 · Commercial intelligence

Market data, credit & risk, legal intelligence, professional and industry datasets. Coverage and confidence, licensed.

Tier 3 · Enterprise private

ERP, CRM, cloud, identity and security systems — connected per customer. The highest-value evidence, unique to each enterprise.

BloombergMARKETS & FINANCIAL DATA
ReutersNEWS & EVENTS
S&P GlobalRATINGS & INDICES
Associated PressNEWS WIRE
NewsGuardSOURCE CREDIBILITY
PolymarketPREDICTION MARKETS
Bright DataWEB & SOCIAL DATA
FREDMACROECONOMIC
Companies HouseSEC EDGARFCA · regulatorsSanctions listsPubMed · arXiv · CrossrefNIST · ISO · W3CWikidata · OpenCorporatesCourt & insolvency registersERP · CRM · cloud · identity~100 mapped providers

The network effect is the moat: more providers improve evidence quality → better evidence strengthens every Trust Object → stronger records attract more enterprises → more enterprises attract more providers. Long-term: certified evidence providers and an evidence marketplace — an ecosystem, not a product.

Target integrations across a mapped ~100-provider ecosystem — capability, not endorsement; subject to licensing. Logos are the property of their owners.

13 · Access

The waitlist is open.

Developer preview

Early access to the API, the MCP server and the sandbox. Build trust-aware applications before public launch.

Enterprise access

Certified issuance, audit-trail retention and governance for regulated teams. Financial services and insurance first.

Individuals & professionals

Work beside AI the way you already do — with a grade on every answer you rely on. Check it, act on it, attach the certificate when you need to show your work.