← Back to Trust Center

AI Data Handling

Last updated: March 25, 2026

CaseProof uses AI to generate chronologies and identify evidentiary gaps. This page describes exactly which models are used, what data is sent to them, and what is retained.

Models Used

  • OpenAI GPT-4o — accessed via the OpenAI API for all AI-generated outputs.

What Is Sent to OpenAI

  • Document text excerpts extracted from uploaded files, used to generate chronologies and gap analyses.
  • Structured prompts constructed server-side that include relevant document text and task instructions.

What Is NOT Sent to OpenAI

  • Raw file binaries (PDFs, images, videos)
  • User account information (name, email, billing data)
  • Authentication credentials or session tokens
  • Any data unrelated to the specific matter being processed

OpenAI Data Retention

OpenAI does not use API inputs or outputs for model training by default. This is governed by OpenAI's API Terms of Service. Zero Data Retention (ZDR) — where OpenAI does not store API request/response data at all — is available on request for Enterprise customers.

To inquire about ZDR for your organization, email legal@case-proof.com.

How Outputs Are Stored

All AI-generated outputs (chronologies, gap analyses) are generated server-side and stored in your Supabase instance only — not retained by OpenAI after the API response is returned. Outputs are associated with your account and are subject to the same access controls and encryption as all other case data.

Human Review Requirement

All AI-generated outputs are labeled "Draft" until reviewed and explicitly signed off by a human. CaseProof does not present AI outputs as final or authoritative. Attorneys are responsible for reviewing and validating all generated content before use in any legal proceeding.