A search result that should not exist
A firm has rolled out your knowledge and drafting product across its corporate team. An associate searches for a clause on a share purchase and gets a helpful result: an extract from a document on another matter. That matter is behind an ethical wall because the firm acts for a party on the other side of a related deal, and the associate is on the wrong side of it.
The associate reports it, as they should. The firm's risk team asks you how your product decides what a user can see. The honest answer is that everyone in the corporate team can see everything the corporate team has uploaded. The rollout is paused.
Why legal tech products miss information barriers
Ethical walls, also called information barriers, are how firms keep information from one matter away from people who must not see it. The firm sets them in its practice management or document management system, or in a dedicated tool that manages walls across systems. Your product usually knows nothing about them.
- Permissions in your product are set by team or by firm, not by matter.
- Search indexes and AI retrieval draw on every document uploaded, with no check against the user's access.
- Walls change over time as new matters are opened, and your product is not told.
- Exports and shared links can carry content to people outside a wall.
- There is no log the firm can use to show who accessed what.
What a breached wall costs
For the firm, a breach of an information barrier can be very serious, with professional and client consequences. For you, one incident can end a rollout at that firm and become a story in a small market. Even without an incident, a product that cannot respect walls fails security reviews, and firms that notice it later may stop using features such as firm-wide search or AI drafting from precedents, which are often the features you sell on.
How we build matter-level access that follows the walls
What we build makes the matter the unit of access in your product, and takes the firm's walls as the source of truth.
- A matter-level permission model: every document, output and index entry in your product belongs to a matter, and a user's access is checked against that matter.
- Wall and access data pulled from where the firm keeps it: their document management system, practice management system or information barrier tool, through its API or a regular sync.
- Checks applied to every route to content: screens, search, AI retrieval, exports, notifications and shared links, not just the main document view.
- Retrieval for AI features filtered before the model sees anything, so a document from a walled matter cannot be used to draft for someone outside the wall.
- Fast updates when walls change, with a regular full reconciliation to catch anything missed, and the product defaulting to deny if wall data is unavailable.
- An access log per matter recording who viewed, searched, exported or had AI outputs drawing on each document, available to the firm's risk team.
| Route to content | How the wall applies |
|---|---|
| Opening a document | Access checked against the matter |
| Firm-wide search | Results filtered to matters the user can access |
| AI drafting from precedents | Retrieval filtered before the model sees content |
| Export or share | Blocked if a recipient is outside the wall |
| Wall data unavailable | Access denied until data is current |
How walls should be set and when they apply is for the firm's risk and compliance team. We make your product apply their decisions reliably and show that it has.
The same search, with walls in place
The associate searches for the same clause. The results come only from matters they have access to, and nothing from behind the wall appears, not even a title. When the firm opens a new matter with a wall the next week, the next sync picks it up. The risk team can open the access log for the walled matter and see who has touched its documents in your product, which is the evidence they need to keep the rollout going.
Could your product cross a wall?
- Permissions in your product are by team or firm, not by matter.
- Search or AI features draw on all documents a team has uploaded.
- Your product does not know about the firm's ethical walls.
- Walls set in the firm's systems do not reach your product when they change.
- You cannot produce a log of who accessed a given matter's documents.