The Legal AI Client Disclosure Survey 2026: What Clients Are Asking, What Firms Are Saying, and How Big the Gap Is

Executive Summary

A significant and largely unacknowledged transparency gap has opened between what corporate legal departments expect from outside counsel on AI use disclosure and what law firms are actually providing. Based on surveys of 247 in-house counsel and 189 law firm partners and legal operations leads conducted in Q1 2026, The Legal Stack finds that 61% of in-house teams have either added or are actively drafting AI disclosure language to their outside counsel guidelines (OCGs), while only 34% of law firms report having a written, firm-wide AI disclosure policy in place. The remainder are operating reactively — responding to client inquiries on an ad hoc basis with no standardized framework, inconsistent detail levels, and, in several documented cases, misleading characterizations of how AI tools are actually being used in matters.

This briefing breaks down the specific data by firm size, client industry sector, and matter type, and concludes with actionable recommendations for both sides of the relationship.

Methodology

The Legal Stack fielded two parallel surveys between January 15 and February 28, 2026.

Survey A — In-House Counsel Panel (n=247): Respondents were drawn from companies with annual revenue above $500 million. The sample included GCs, deputy GCs, and senior legal operations managers across financial services (22%), technology (19%), healthcare and life sciences (17%), manufacturing (14%), energy (11%), and retail/consumer (17%). Respondents were asked about whether they had requested AI use disclosure from outside counsel, what specific information they sought, and whether firm responses met their expectations.

Survey B — Law Firm Partner and Legal Ops Panel (n=189): Respondents included equity and non-equity partners, chief operating officers, and legal operations directors at Am Law 200 firms (41%), midsize firms of 50–200 attorneys (33%), and boutique or specialty firms under 50 attorneys (26%). They were asked about existing disclosure policies, client inquiry volume, and internal governance structures for AI deployment.

All data was collected anonymously. Firm and client names referenced in qualitative findings were disclosed voluntarily and have been verified independently.

Key Finding 1: Client Demand Is Running Well Ahead of the Market

Sixty-one percent of in-house respondents have added or are drafting AI disclosure requirements to their OCGs. Among technology and financial services companies, that figure rises to 74% and 69% respectively — sectors where data governance maturity and regulatory sensitivity to third-party AI exposure are highest.

Among the 39% without formal OCG language, the majority — 58% of that subset — reported that AI disclosure is on their agenda for the next contract cycle. In practical terms, this means that within 12–18 months, the clear majority of large corporate legal departments will have some form of contractual AI disclosure requirement in place.

What is being requested is increasingly granular. When asked which specific disclosures they want from outside counsel:

79% want to know whether AI tools are being used on their matter at all
71% want to know the names of specific tools (e.g., Harvey, CoCounsel, Microsoft Copilot, Luminance, Westlaw AI)
68% want to understand human review procedures — specifically, whether AI outputs are being reviewed by a licensed attorney before use
54% want information about training data policies, including whether client data could be used to train or fine-tune models
47% want billing transparency — whether AI efficiency gains are being reflected in fees charged
31% want to know whether the firm's AI tools comply with jurisdiction-specific bar guidance (a figure that nearly doubles for matters involving EU-facing work, at 59%)

Financial services clients lead on nearly every category, a reflection of their own obligations under frameworks like the EU AI Act and SEC guidance on algorithmic accountability.

Key Finding 2: Most Firms Are Not Ready for What Clients Are Asking

Only 34% of law firms surveyed have a written, firm-wide AI disclosure policy. The breakdown by size is stark:

Firm Size	Written Policy in Place	Ad Hoc Only	No Policy, No Process
Am Law 50	61%	29%	10%
Am Law 51–200	38%	44%	18%
Midsize (50–200 atty)	22%	41%	37%
Boutique (<50 atty)	11%	38%	51%

Am Law 50 firms — many of which have invested heavily in Harvey deployments, custom LLM integrations, and dedicated AI governance committees — are better positioned, though even among this cohort, 29% are still operating without standardized disclosure language, relying instead on partner discretion.

The midsize and boutique segments represent the most acute risk. These firms increasingly use AI tools (particularly commoditized offerings like Clio AI, Lexis+ AI, and Copilot for Microsoft 365) but lack the governance infrastructure to document and communicate their use coherently. One legal ops director at a 90-attorney regional firm described the situation plainly: "We use these tools constantly. If a client asked me tomorrow to enumerate every AI touchpoint on their matter last year, I genuinely couldn't."

Key Finding 3: The Mismatch Is Sharpest on Three Specific Issues

Tool identification. Clients want tool names. Firms resist, citing competitive confidentiality, workflow complexity, and the pace at which their tool stacks evolve. Only 29% of firms say they currently disclose specific product names in response to client inquiries — though 71% of clients consider this a baseline expectation.

Data handling and training policies. This is the single area where the gap generates the most friction. Fifty-four percent of in-house respondents want training data disclosures. Only 18% of firms report being able to provide this information confidently. The problem is not unwillingness but structural ignorance: most firms have not obtained written confirmation from vendors like Harvey or Microsoft about whether enterprise-tier contractual protections fully prevent client data from being used in model training or fine-tuning. Harvey's enterprise agreements, for instance, contain explicit prohibitions on using client data for training, but fewer than half of the firm respondents who use Harvey could affirmatively confirm they had reviewed or retained those contract provisions.

Billing transparency. This is the least-addressed and potentially most legally consequential gap. When AI tools reduce document review time by 60–80% — figures consistent with published benchmarks from firms using Luminance and Relativity aiR — the question of whether clients are being billed for time that was never spent is not theoretical. Forty-seven percent of clients want efficiency disclosures. Only 12% of firms have any internal policy that addresses how AI-generated time savings should affect billing. This is fertile territory for future fee disputes and, in more aggressive scenarios, professional responsibility complaints under Model Rule 1.5.

Key Finding 4: Matter Type Drives Disclosure Intensity

Client expectations vary significantly by matter type:

Matter Type	% Clients Requiring Disclosure	Primary Concern
M&A due diligence	78%	Data handling, confidentiality
Litigation / e-discovery	74%	Accuracy, human review, court rules
Regulatory / compliance	71%	Jurisdictional AI rules, data residency
Commercial contracts	52%	Efficiency billing, accuracy
Employment matters	48%	Confidentiality of sensitive HR data
General outside counsel work	41%	General oversight

M&A and litigation clients are applying the most pressure, driven by the sensitivity of documents being processed and — in litigation — the expanding body of court rules requiring disclosure of AI use in filed documents. Standing orders from judges in the Northern District of California, the Southern District of New York, and the Fifth Circuit have already created de facto disclosure obligations in federal practice that have yet to be consistently integrated into firm-level client communication protocols.

Recommendations

For Law Firms:

Draft a written AI disclosure policy now, even if imperfect. The Am Law 50 cohort demonstrates that having a policy — even a preliminary one — meaningfully improves client satisfaction scores. A one-page framework that identifies which tools are deployed firm-wide, what human review steps are standard, and how data is handled is sufficient to begin. Update it quarterly.
Audit your vendor contracts for data training protections. Obtain written confirmation from Harvey, Microsoft, Lexis, Thomson Reuters, and any other AI vendors that client data is excluded from model training under your enterprise agreements. Retain these documents. Be able to produce them.
Address the billing question before clients do. Develop an internal policy — in consultation with your ethics counsel — on how AI efficiency gains affect hourly billing. Consider proactive disclosure as a client relations differentiator rather than a liability.
Create a disclosure coordinator function. Designate a legal ops or client relations lead as the single point of contact for AI disclosure inquiries. Reactive, partner-by-partner responses are the source of most inconsistency and client dissatisfaction.

For Legal Departments:

Update your OCGs now, not next cycle. Model language is available from the ACC, the Coalition of Law Company Counsel, and individual GC networks. Waiting for the perfect framework means another 18 months of uninformed AI use on your matters.
Ask specifically about billing. Do not limit disclosure requests to tool usage. Ask directly whether AI efficiency is being reflected in fees. Document the response.
Tier your disclosure requirements by matter type. Reserve your most detailed requirements for M&A, litigation, and regulatory matters. Applying the same standard to a routine contract review creates friction without proportionate value.
Accept that disclosure is an ongoing obligation, not a one-time certification. AI tool stacks at law firms are changing monthly. Build annual re-certification into your outside counsel relationships.

Conclusion

The disclosure gap documented here is not primarily a technology problem or a goodwill problem. It is a governance and communication problem — and it is solvable. The firms and legal departments that close this gap in 2026 will be better positioned for the more complex regulatory environment ahead, including anticipated ABA Model Rule amendments addressing AI competence and the expanding reach of the EU AI Act into transatlantic legal practice. The survey data suggests that most clients are not yet punishing firms for disclosure failures. That window will not remain open indefinitely.

Survey data and methodology available to subscribers on request. The Legal Stack does not accept sponsored research funding. All survey instruments were designed and administered independently.