A vendor-agnostic overlay above whichever legal AI tool a firm or in-house function already uses. We do not review the model's legal reasoning. We evidence that its use was authorised, privilege-safe, reviewed by a named human, and reconstructable afterwards.
Independent assurance layer. Cabier does not provide legal advice and does not supplant the firm's own supervisory obligations.
Each module is a vertical projection of an asset already running on the platform. No parallel store, no second evidence chain.
AI Evidence Vault
Unified Evidence Vault
Prompt, retrieved sources, output, reviewer and disposition captured independently of the vendor console, with an audit identifier per record.
Prompt Governance Manager
Sovereign AI Governance OS — prompt governance
Version-controlled prompt templates for high-stakes work, with privilege and confidentiality tags applied at the asset level.
Decision Log
Unified Audit Trail
An ordered record of who asked, what the model returned, who reviewed it and what left the firm.
A scored governance position across policy, competence, logging, privilege, disclosure, prompt control, review chain and reconstruction — refreshed continuously, not annually.
A separate measure of whether privileged material can leave the boundary at all, tested rather than asserted.
Where unreviewed or unlogged model use is accumulating across practice groups, surfaced before it becomes a claim.
A scoped attestation the firm can hand to its own client, naming period, scope and stated limitations.
Eight questions across the dimensions a challenge would actually test. Answers stay in your browser — nothing is submitted, stored or transmitted.
01 · Policy
Is there an approved written policy stating which AI tools may be used on client matters, and for what?
02 · Competence
Have fee earners been trained on the policy and recorded as having completed it?
03 · Logging
Are prompts and model outputs captured and retained outside the vendor's own console?
04 · Privilege
Is there a control that prevents privileged or confidential client material entering an unapproved model?
05 · Disclosure
Is there a defined position on whether and how clients are told AI was used on their matter?
06 · Prompt control
Are prompt templates for high-stakes work reviewed and version-controlled rather than authored ad hoc?
07 · Review chain
Is a named human reviewer recorded against every AI-assisted work product that leaves the firm?
08 · Reconstruction
Could the firm reconstruct, on request, how a specific AI-assisted output was produced?
Answer the questions above to see an indicative position.
Days 1–15
Scope and obligation mapping
Practice groups in scope, tools in use, obligation set fixed, evidence types agreed.
Days 16–45
Capture and prompt governance
Independent evidence capture stood up; high-stakes prompt templates brought under version control.
Days 46–75
Control testing
Privilege boundary, review chain and reconstruction tested against live matters.
Days 76–90
Position and attestation
Maturity position issued, gap remediation sequenced, certificate scope agreed.
Scope, residency and retention are set per engagement. Commercial terms are quoted against scope.
Start a scoping conversation