Governing the whole response, not just its citations.

The request, and the matter it belongs to

You draft, research and organize in the KeepWise workspace. Chats are organized under clients and matters, so their notes, evidence and AI-generated documents stay with the work they belong to.

Owners and attorneys can see every client and matter in the firm. Roles (owner, attorney, paralegal, staff) decide what each person can do; an owner or attorney can give a paralegal or staff member access to a client and its current matters from the client's settings.

If a client or matter is under a legal hold, nothing under it can be deleted or moved until the hold is lifted.

Held until the checks finish

The AI's draft is held in a buffer while automatic checks run against it: does each cited case exist, does quoted language appear in that opinion (both checked against CourtListener's case-law database), and does the response claim to have done something the AI has no tool to do.

Deterministic checks that fail rewrite or block the response before release. A model-based behavioral review also runs; it is recorded in the audit trail rather than changing the response.

These checks run on every plan, automatically. No one has to remember to run them under deadline.

Results you can see, next to the answer

Verification results appear inline, moments after the response text. If the citation check can't run, the response's citations show no verdict and the response says they could not be checked.

A person makes the consequential calls

Actions with real side effects don't execute automatically: they're proposed, then require a person's explicit approval before anything runs.

With Verifier Pro, whether a case actually supports the point it's cited for is routed to an attorney at your own firm to confirm. That judgment is never auto-cleared, and the confirmation is logged.

Lifting a legal hold needs a second attorney or admin to approve. If you are the only one, the lift waits 72 hours, and you can cancel it until then.

A record that shows tampering

Audit events are written to a hash-chained ledger: each record's hash includes the one before it, so editing or deleting a past entry breaks the chain. The chain's starting value is kept outside the database.

Approval decisions are kept in a separate append-only record that standard database roles cannot edit or delete; that record is not hash-chained.