Trust Infrastructure

Trust = decision + receipt + auditability.
This page documents our infrastructure commitments. Artifacts included.

What we mean by trust

Trust is measurable. It's a permissioned decision under uncertainty - with proof.

Decision

Yes / Conditional / No. Clear, actionable, enforceable.

Receipt

Proof of what was decided, when, and why. Tamper-evident.

Auditability

Searchable, versioned, transparent. For you and your auditors.

Receipts & Audit Trails

Every decision emits a receipt. Every receipt is searchable, exportable, and auditable.

Receipt Contents

Receipt ID	Unique identifier (e.g., `rcpt_7f3k9x2m`)
Decision	Yes / Conditional / No
Confidence Tier	Low / Medium / High
Reason Category	Non-sensitive explanation
Timestamp	UTC, ISO 8601 format
Flow ID	Which flow triggered this decision
Policy ID + Version	Which policy was applied
Signals Used	Categories only (not raw PII)
Override Log	If human intervention occurred
Audit Hash	Tamper-evident integrity check

Search by: Flow ID, Policy ID, Receipt ID, Time Range, Decision Type

Policy Engine

Policies are versioned, testable, and rollback-ready. No surprises.

Versioning

Every policy change creates a new version. Full history preserved. Receipts reference exact version used.

Rollback

One-click rollback to any previous version. New bad policy? Undo in seconds.

Change Logs

Who changed what, when, and why. Auditors love this.

Sandbox Mode

Test policy changes against real traffic without affecting production decisions.

Data Handling

We minimize data by default. Retention is configurable. You control what stays.

Data Retention Matrix

Data Type	Default Retention	Configurable?
Decision receipts	90 days	Yes
Signal categories	90 days	Yes
Raw input data	Not stored	N/A
Policy versions	Indefinite	Limited
Override logs	1 year	Yes

Data minimization

We don't store what we don't need

Redaction support

Sensitive fields can be redacted

Access controls

Role-based access to data

Human Review & Overrides

Conditional decisions route to your review queue. Overrides are logged and produce their own receipts.

Conditional Routing

Conditional decisions enter review queue
Configurable escalation rules
SLA tracking on review time

Override Requirements

Reason required (dropdown + notes)
Override produces its own receipt
Full audit trail preserved

Note: Override receipts include: who overrode, when, reason category, and resulting decision.

Abuse Controls

Controls to prevent abuse and protect system integrity.

Rate Limiting

Configurable per-flow, per-user, and global limits. Prevent enumeration attacks.

Repeated Actor Controls

Detect and flag repeat offenders across sessions and identities.

Monitoring & Alerting

Real-time monitoring. Configurable alerts for unusual patterns.

Bot Detection

Identify automated attempts to bypass verification.

AI Governance

How our models work, what they don't do, and how you control them.

What the model does

Classification of document types and authenticity signals
Anomaly detection for unusual patterns
Risk scoring based on configurable thresholds
Signal aggregation into decision categories

What it doesn't do

No raw PII exposed in receipts
No biometric claims unless explicitly enabled
No autonomous decisions without policy constraints
No training on your data without consent

Evaluation & Monitoring All systems nominal

Metric	Description	Threshold	Current	Last checked
`false_positive_rate`	Legitimate requests incorrectly flagged	< 2%	0.8%	Continuous
`false_negative_rate`	Risk events missed by the model	< 1%	0.3%	Continuous
`model_drift_score`	Performance degradation over time	< 0.05	0.02	2026-01-20
`latency_p99`	Decision response time	< 500ms	247ms	Real-time
`override_rate`	Human overrides of model decisions	< 5%	1.2%	2026-01-21

Metrics are aggregated across all flows. Per-flow breakdowns available in your dashboard.

Threshold controls

Set confidence thresholds for Yes/Conditional/No

Human-in-the-loop

Conditional routes to review; overrides logged

Fallback behavior

You control hold/allow when model is unavailable

Limitations: No model is perfect. We provide confidence tiers, not certainty. Conditional exists specifically to handle uncertainty with human judgment.

Reliability & Incidents

Trust means telling you when things aren't perfect.

Status Page

Real-time system status. Last 90 days uptime. Incident history.

View Status

Incident Process

Detection + status page update
Initial investigation
Mitigation + communication
Resolution + post-mortem

When we're down

Graceful fallback: You configure default allow/deny behavior
Kill switch: You can bypass Werify entirely if needed
Latency expectations: Decisions typically return in <100ms

Compliance Posture

We only list controls we have now. Roadmap items are labeled clearly.

Available Now

Data encryption at rest + in transit
Role-based access controls
Audit logging
Data retention controls
SOC 2 Type II In Progress

Planned

ISO 27001 certification
GDPR data processing addendum
HIPAA BAA (healthcare customers)
PCI DSS compliance

Need specific compliance documentation? Contact our security team.

Artifacts Library

Downloadable artifacts for your security review and integration.

Download the Trust Pack

Forwardable to security and procurement - no call required.

Browse Trust Pack

Sample receipts (4) Retention matrix Architecture diagram Failure modes

Updated: 2026-01-22 • Version: v1.0

Sample Receipts

4 receipt types (Identity, Documents, Access, Transactions)

Identity Docs Access Txns

Architecture Diagram

System architecture and data flow

Download PDF

Data Retention Matrix

Full retention policy details

Download JSON

Policy Templates

Starter templates for common use cases

Browse Templates

Failure Mode Behavior

What happens when we're down

Download PDF

Security Whitepaper

Detailed security controls overview

Download PDF

Questions about our trust infrastructure?

Our security team is happy to walk you through the details.

Talk to security Read the docs

Graceful fallbacks · Kill-switch · Configurable retention · View status