Continuous AI Assurance scoring across every model, every regulation, every day. Your data never leaves your environment.
Local-first · Air-gapped capable · Apache-licensed open models
VERACIS
ENTERPRISE AI ASSURANCE INDEX · Q1 2026
"Defensible posture, material OPI gaps."
Upload model risk policies, validation reports, governance memos. Veracis parses, chunks, and indexes everything locally.
A frozen, versioned methodology maps evidence to CBUAE dimensions and produces RAG scores with citations to source paragraphs.
Interactive chat over findings. Exportable Dimension Status Report PDF. Every score traceable to the rule that produced it.
The Veracis AI Assurance score is deterministic and methodology-versioned, computed across five weighted dimensions. Together they total 100% and feed the board-grade Dimension Status Report.
Board accountability, Model Oversight Committee operation, three-lines-of-defence, escalation paths.
Model inventory completeness, tiering integrity, lifecycle gates, ownership, validation independence.
Mapping to CBUAE, EU AI Act, ISO 42001, NIST AI RMF — clause-by-clause coverage, gap remediation timelines.
Performance monitoring, drift detection, override governance, incident response, change management.
Documentation completeness, validation depth, audit-trail traceability, methodology versioning.
Total weight: 100% · methodology v1.2 · deterministic, regulator-defensible
Red, Amber, Green across CBUAE Model Management dimensions — with the underlying evidence one click away.
Ask questions of the findings. Every answer cites the source paragraph, the dimension, and the rule that fired.
Every report stamps the methodology hash. When the rules change, the lineage is preserved — and so is your audit trail.
One-click export of a presentation-ready PDF for risk committees, internal audit, and supervisory reviews.
The same evidence base — read differently by each role that has to answer to a regulator.
Enterprise score across countries and models. Drill from a 67 into the dimension that's bleeding.
Open findings, model inventory with evidence on add, reasoning trace, two reporting packs ready to send.
Plain-English status on the models running your customer journeys. No model-risk jargon required.
Every finding mapped to a clause. CBUAE 4881 fully indexed; EU AI Act material-changes view built in.
Versioned methodology hash on every report. Replay any score on any historical evidence set.
One-page board pack. The same number across HOM, CRO, and Audit — no version drift.
Read each row left to right. The contrast is the point.
Veracis runs entirely on your infrastructure — laptop for evaluation, on-prem for production. Open-weight models. No cloud round-trips. No telemetry. No vendor lock-in.
Live demo against bundled CBUAE governance data — no install, no upload, no signup forms.
Local-first · Air-gapped capable · Your data never leaves