A site-level wall leaks through search, email and personal storage
None of these are configuration errors in isolation. Together they mean the wall exists in the site and nowhere else.
Most firms have a conflicts policy and a matter site structure. Very few have tested whether the wall exists anywhere other than in the site.
A firm creates a site per matter, restricts membership, and considers the ethical wall built. The problem is everything around the site: the email thread in personal mailboxes, the document saved to a partner's own storage, and the search result that returns a title and a snippet to somebody who cannot open the file.
This paper argues that an ethical wall built only at the site level is not a wall, sets out the four layers that must agree for it to be one, and proposes a test firms can run themselves before their risk partner has to.
If you read nothing else, read these. The analysis that follows sets out the evidence for each.
None of these are configuration errors in isolation. Together they mean the wall exists in the site and nowhere else.
It reveals the existence and subject of a matter without revealing the file, which is a confidentiality event in its own right.
A dedicated directory attribute changes only when somebody deliberately changes it. Department changes for reasons unconnected to conflicts.
The work required is identical whether or not you deploy an assistant. What changes is how quickly the gap is found and by whom.
A defensible ethical wall in Microsoft 365 is not one control. It is four, and they have to be designed together rather than added in sequence.
Information barriers provide policy-level segmentation preventing communication and collaboration between defined groups, enforced across Teams, SharePoint and OneDrive. Access design drives matter site membership from the practice management system rather than maintaining it by hand. Sensitivity labels apply protection that travels with the document when it leaves the site. Retention aligns to matter lifecycle rather than to a calendar date, so closed matters actually close.
Firms that implement one of these and consider the wall built are the norm rather than the exception.
Three characteristics compound. Partnerships resist mandated process, so configuration that fights how partners work is worked around. Matter teams form and dissolve continuously, so membership drifts faster than in a corporate structure. And the content is unusually sensitive, so the consequence of a leak is professional rather than merely commercial.
Microsoft is investing heavily in making content discoverable and answerable by agents — automated metadata, stale content detection, assistants that reason over libraries. For a firm with a properly built wall this is straightforwardly useful. For a firm whose wall exists only at the site level, it is a considerably faster way to discover the gap.
The remediation work is the same either way. The question is only whether the firm does it deliberately or after an incident.
The single most useful hour in this work is a deliberate attempt to breach your own wall.
Take an account on the wrong side of a barrier and try, methodically, to reach the material: search, direct URL, a link forwarded by a colleague, a document opened on a personal device, an export to a spreadsheet. Write down what worked.
In our experience the search result snippet surprises risk partners most often. It is not a permissions failure in the conventional sense — the file cannot be opened — and it still discloses that the firm is acting on a matter and what the matter concerns.
Every paper in this series ends with a framework you can run internally. We would rather you used it and reached your own conclusion than took ours on trust.
Five stages including the test. Firms that stop after stage two have a site, not a wall.
A dedicated directory attribute populated for 100% of users. Unpopulated users sit outside every policy.
Barriers defined and membership signed off by the risk partner before enforcement.
Sensitivity labels so protection travels with the document beyond the site.
Retention triggered by matter closure, so closed matters leave the discoverable estate.
Attempt to defeat your own wall, in writing, before anybody else does.
The same argument lands differently across an executive team. These are the three versions worth separating.
Microsoft's own documentation for the product behaviour described above. We would rather you verified the basis than accepted our summary of it.
On these references: each entry names a Microsoft Learn article or documentation area by title, because deep links change while titles are stable. Searching the title on learn.microsoft.com will reach the current version. Where we have cited a figure or a product behaviour, it is Microsoft's statement rather than ours; where we have given a number of our own it is labelled as such in the text.
We will run a scoped information barrier review against one practice group, including the deliberate breach test, and give you the findings in writing whether or not you take the remediation further with us.
For a firm holding client confidences across jurisdictions, residency is a professional obligation before it is a technical constraint.
A control that depends on partners behaving differently from how they demonstrably behave is not a control — and in a negligence context it is worse than none.
Describe the situation in your own words.