Governing enterprise AI agents, or a law firm's own AI use?
Roval sits in the same broad AI-governance category and promises audit-readiness, but it's aimed at governing enterprise AI agents. Candor is narrower on purpose: it governs how a law firm's own attorneys and staff use outside AI tools on client matters. Here's the split.
| Feature | Roval | Candor |
|---|---|---|
| Who it's built for | Enterprises deploying AI agents | U.S. law firms, specifically |
| What it governs | Autonomous AI agents an organization runs | How a firm's people use outside AI tools on matters |
| Framing | Enterprise agent oversight and audit-readiness | Bar rules and ABA Formal Opinion 512; a carrier-renewal report |
| Confidentiality screening | Not its focus | Screens client details before text reaches a model |
| Citation checks | Not its focus | Flags cited cases that may need a second look (CourtListener) |
| Attorney sign-off + matter log | Not its focus | Per-matter use log with recorded attorney review |
| Record for a malpractice carrier | Not its focus | Append-only, hash-chained, carrier-ready |
| Runs on | Enterprise deployment | The firm's own connected model; one-click browser pairing |
This compares stated focus, not quality. If your question is "can we prove our own attorneys used AI responsibly on a matter," that's the Candor side.
A firm's risk isn't an agent it deployed. It's a brief nobody checked.
Agent governance answers "is this autonomous system behaving." A law firm's exposure is different and human: a paralegal pasting client facts into ChatGPT, or a citation that doesn't exist making it into a filing. The artifact that answers for that is a per-matter record with screened inputs, verified citations, and an attorney's sign-off, which is exactly what Candor produces.