Skip to content
OnTrial
AI Governance

AI that accelerates the work — not the risk.

The firms OnTrial serves carry duties to their clients — privilege protection, work-product integrity, conflicts management, and increasingly, contractual outside counsel guidelines that govern vendor AI use. OnTrial is built to meet those duties and to give you the documentation to prove it.

This page covers our AI governance posture specifically. For the broader security posture — encryption, access control, infrastructure — see Data Security.

Core principles

Four commitments. Non-negotiable.

  • No client data trains any model

    Your matter information is never used to train, tune, or fine-tune AI systems — internal or third-party. This is an architectural commitment, not a policy preference.

  • Zero retention in the processing layer

    Case data flows through isolated infrastructure and is purged from the processing environment within 24 hours. Long-term storage is governed by your retention preferences and applicable law.

  • Human-in-the-loop on every finding

    AI accelerates the investigative pipeline. A veteran or LEO case manager reviews and an analyst verifies every finding before delivery. AI does not sign off on output.

  • Defensible for outside counsel guidelines

    Built to satisfy the AI-use, data-handling, and vendor-management provisions that increasingly appear in OCG and engagement letters. Procurement-reviewable documentation on request.

How AI is used

AI widens the funnel. Analysts decide what enters a report.

The technology shows up at four stages of the investigative pipeline. In every one, a human is responsible for what reaches the client.

  • 01

    Collection acceleration

    AI-assisted retrieval across permitted sources lets a single matter pull from substantially more inputs than manual collection alone — without compromising sourcing discipline.

  • 02

    Pattern recognition & triage

    Machine analysis surfaces signals, anomalies, and connections an analyst then verifies. The technology widens the funnel; the analyst decides what enters a report.

  • 03

    Document review

    Large unstructured corpora — filings, transcripts, social archives — are pre-processed for relevant excerpts. Final extraction is analyst-driven against the matter's evidentiary standard.

  • 04

    Workflow and case management

    AI assists with intake structuring, scope documentation, and case-management hygiene. None of it touches substantive findings without analyst review.

Auditability

Every finding has a documented path.

When a report supports a deposition, a motion, or a trial decision, the methodology has to hold up. Our case-management layer logs every step so the same answer comes back regardless of who asks: how was this finding reached?

  • Every step of the investigative process is logged and timestamped.

  • Chain of custody is documented from intake to delivery.

  • Reports are designed to support deposition, motion practice, and trial use.

  • Activity records are retained per applicable retention rules and produced on subpoena, court order, or client request.

Outside counsel guidelines

Built for the AI provisions in modern OCG.

The AI-use, data-handling, and vendor-management provisions that now appear in outside counsel guidelines and engagement letters are increasingly specific. OnTrial maps to them by design.

  • Vendor AI use disclosure

    OnTrial's AI use is disclosable in writing to the engagement client and their counsel on request.

  • Data segregation

    Client matter data is processed in segregated infrastructure separated from any model provider's training data pipeline.

  • No cross-client data reuse

    Findings, intermediate artifacts, and case data from one engagement are never reused as inputs to another engagement.

  • Human accountability

    A named veteran or LEO case manager is accountable for every engagement from intake through delivery.

  • Documented audit trail

    Investigative steps, sources consulted, and verifications performed are logged in a form that supports defensible reporting.

Procurement-reviewable documentation, on request.

AI-use disclosures, vendor-management documentation, and the underlying Data Security posture are available to clients and their procurement teams. Reach out and we'll put a packet together for your matter.