The Perimeter Has Shifted: Findings from the 2026 State of Data Protection in Legal Report

Law firms don't sell hours. They sell trust.

Everything a firm monetizes, including billing power, client retention, panel position, referrals, sits on top of one belief: that the client's information is safe with you. Confidentiality isn't a feature of the service. It is the product. The economic value that confidence generates is what we call the Trust Premium, and it has quietly become the most exposed asset in the legal sector.

That premium is under structural pressure — not because firms stopped protecting client data, but because the place where protection has to happen moved. Legal work now requires sensitive documents to travel: to opposing counsel, client portals, due diligence platforms, AI tools. The perimeter has shifted to wherever the document goes, while most governance frameworks were built for a world where data stayed inside the building.

What the Research Found — and How We Found It

How wide is that gap, and where does it actually show up? That's the question behind our 2026 State of Data Protection in the Legal Sector report — a strategic risk analysis of more than 200 prominent U.S., Canadian, and UK law firms examining governance gaps, AI exposure, and the document perimeter problem.

The research was built on open-source intelligence gathering — firm websites, technology partnership announcements, legal industry databases, regulatory filings, breach notifications, and media records. No firm provided internal data, which means the picture you're about to see is almost certainly a conservative floor for actual exposure.

One more piece of context matters before the findings. The Trust Premium was previously maintained almost entirely through reputation. That era is ending. Clients have moved from assuming trust to actively verifying it, and the verification is concrete: security questionnaires, procurement requirements, and outside counsel guidelines with explicit data protection clauses. That shift from reputational assurance to technical verification is the structural change that makes everything below relevant.

Below are the four takeaways from the report, and how legal firms can adjust moving forward.

Four Takeaways from 200+ Law Firms

1. The perimeter didn't die. It shifted.

The governance frameworks most firms have built were designed around one core assumption: data stays inside the building. For a long time, that was reasonable. Secure the walls, secure the systems inside them. All set.

That's not how legal work operates anymore, because documents have to move. Whether it’s to opposing counsel, expert witnesses, client portals, due diligence platforms, or now AI tools, data is always on the move. Keep in mind - this isn’t a security failure. It’s the new reality of the legal firm. That's not a security failure - it's just the new way of work.

The document management system (DMS) isn't the weak point either. However, they are built to govern documents inside a controlled environment, and they do it well. But the moment a document leaves that environment, the access controls, retention policies, and classification tags stay behind. This means the container remains secure while the asset itself isn’t. 

What this tells us is that the perimeter now travels with the document while the controls stay at home.

Why is perimeter security no longer enough for law firms? Perimeter security assumes sensitive data stays inside firm systems. Legal work requires the opposite: documents move constantly to opposing counsel, clients, deal platforms, and AI tools. Once a file leaves the DMS, traditional access controls and classification tags stop traveling with it, leaving the firm's most sensitive work product unmanaged.

Why is perimeter security no longer enough for law firms?

Perimeter security assumes that sensitive data remains within the firm's systems. There is some irony here because legal work requires the opposite. Documents and data are always on the go, whether it’s to opposing counsel, clients, deal platforms, or AI tools. And once a file leaves the DMS, traditional access controls and classification tags stop traveling with it, leaving the firm's most sensitive work product unmanaged.

2. Exposure is structural — not a sign of negligence.

Attackers don't target law firms because they're an easy target. Rather, they target them because of what firms hold.

A sealed acquisition agreement at the company level is one deal. At the firm level, it's every deal — across every client, every sector, every jurisdiction the firm touches. A single M&A partner might have twenty live matters running at once; none of them public, and all of them in the same environment. This means an attacker who gets in doesn't have one secret - they have a hedge fund's worth of non-public market intelligence.

The research bears that out. Across 200+ firms, exposure wasn't a function of weak security programs — it was a function of the work itself: the most sensitive practice areas, clients operating under HIPAA, GDPR, SEC, and FCA frameworks whose own regulators require them to hold outside counsel to the same standards, matters moving across borders and organizational boundaries by design. Even the best-positioned firms in the study carried meaningful structural exposure, with no governance failure to explain it. That risk walks in the door with the engagement letter.

None of this stems from firms doing something wrong — some exposure simply comes with the territory. How that exposure is distributed across the sector and which factors drive it are what the report breaks down in detail.

3. The Trust Premium is now actively priced.

Clients are asking the same question attackers already answered: what does this firm hold, and how well is it protected?

87% of corporate clients say security posture directly influences which outside counsel they select. 75% now require rigorous security questionnaires as part of procurement. And what they're really asking isn't "are your walls secure?" It's "what happens to our data across the life of the matter?" — when it's drafted, shared with opposing counsel, moved through deal platforms, and reviewed by AI tools.

That makes security posture a commercial variable. A breach, or a questionnaire a firm can't answer with confidence, isn't just a forensic problem. It's a business development problem: tightened requirements, matters quietly moving to other firms, RFP shortlists you fall off without being told why. The financial cost of an incident is recoverable. The client's confidence is harder to replace.

The Trust Premium firms have always monetized is now something they have to demonstrably earn — matter by matter, across the full lifecycle of the data.

4. Shadow AI is the governance gap firms can't see.

When a document leaves by email, there's at least a trail — a sent item, a log entry, something. When an attorney opens a consumer AI tool and pastes in a client document, there's nothing. No DMS entry, no access log, no alert, and no way for the firm to know it happened.

What is shadow AI in law firms? Shadow AI is the use of AI tools outside a firm's sanctioned, governed environment — an attorney pasting client material into a consumer chatbot, for example. It leaves no DMS entry, access log, or alert, making it invisible to firm governance and a growing source of confidentiality and privilege risk.

What is shadow AI in law firms?

Shadow AI is the use of AI tools outside a firm's sanctioned, governed environment — an attorney pasting client material into a consumer chatbot, for example. It leaves no DMS entry, access log, or alert, making it invisible to firm governance and a growing source of confidentiality and privilege risk.

The numbers show how wide this gap runs. 33% of law firm professionals use generative AI at least once per day. Yet 44% of firms had no formal AI governance policy as of 2025, and only 15% have technical controls that block unauthorized AI access. Across industries, 77% of employees regularly paste data into generative AI tools — 82% of it through unmanaged personal accounts — and 40% of uploaded files contain PII or PCI.

The stakes sharpened in February 2026, when a U.S. District Court ruled in US v. Heppner that communications made on a publicly available AI platform were not protected by the attorney-client privilege because the platform's privacy policy created no reasonable expectation of confidentiality. Unsanctioned AI use isn't just a governance problem. It's potentially a privilege waiver.

The issue isn't that attorneys are reckless. Most are trying to do their jobs more efficiently. But a policy that says "don't use unsanctioned AI tools" doesn't close the gap, because a policy isn't a control.

What This Means for Firms

The exposure data is a symptom of an architectural problem. Some exposure is structural and comes with the territory — but the gap between what clients now verify and what firms can actually prove is widening, and most current governance models weren't built to handle it.

That's not a reason for alarm. It's a reason to ask a precise question: does the architecture underlying your security program reflect how your data actually moves? If your governance ends where the DMS does, the answer is no, and your clients' security questionnaires are starting to surface that.

Governance Must Travel with the Data

If the perimeter is the document itself, protection has to travel with the file. That's the architectural principle behind Confidencial's approach: selective encryption and access controls that follow the document after it leaves the DMS, the email, and the firm's boundary. Confidencial keeps matter content controlled and auditable wherever it moves, including into AI workflows. For firms across the U.S., Canada, and the UK facing client audits, outside counsel guidelines, and rising AI governance expectations, that's the difference between asserting protection and proving it. You can see how this maps to the patterns in the research on our legal sector solutions page.

Read the Full Report

The perimeter has shifted, but for most firms, the governance model hasn't. The full 2026 State of Data Protection in the Legal Sector report covers the complete scoring methodology, all eight risk dimensions, the sector-wide findings, and what closing the gap requires — read it here.