N
NetDocuments
NetDocuments is a cloud document and email management platform for law firms, corporate legal departments and public sector legal teams, used inside Microsoft Word, Outlook and Teams, a web application and mobile apps. Its AI suite, ndMAX, includes a Legal AI Assistant that answers questions across one or hundreds of documents, AI Search, Tabular Review, an App Builder for creating AI apps without a developer, and a library of ready-to-use legal apps in the ndMAX Studio covering tasks such as automatic profiling, NDA analysis, contract risk analysis, playbook review and deposition preparation; Smart Answers, which began rolling out to ndMAX Enterprise customers in March 2026, answers questions in natural language from the firm's own repository and matter history with citations. The generative features run on Microsoft Cognitive Services and Azure OpenAI, and the company states that inputs and outputs are not kept beyond producing the answer and are not used to train the model, and that its agreement with Microsoft exempts customer content from abuse monitoring. AI operates inside the firm's existing permissions and ethical walls, and Model Context Protocol connections let outside AI tools including Claude, ChatGPT, Microsoft Copilot, Harvey and Legora reach that content without it leaving the platform. NetDocuments Software, Inc. is headquartered in Lehi, Utah and contracts through separate entities for the Americas, for the United Kingdom, Europe, Africa and the Middle East, and for Australia, New Zealand and Asia; customers choose the region their data is stored in, with sign-in points for the United States, United Kingdom, Germany, Australia and Canada, and the company states it supports more than 7,000 law firms, corporate legal departments and public sector organisations. In August 2026 it published the Legal Context Engineering Benchmark, which reports measured answer-quality scores and the cost of a correct answer across 300 questions on ten real matters, together with the method it used.
Capability grades
All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
The models are the engine of a core capability layered on a product that stands entirely without them. NetDocuments is a cloud document and email management system that has been sold for twenty-five years; the ndMAX suite adds a Legal AI Assistant, AI Search, Tabular Review, an App Builder and a library of ready-to-use legal apps, and Smart Answers, which began rolling out to ndMAX Enterprise customers on 31 March 2026, answers questions from the firm's repository with citations. The vendor's own framing is that it brings AI to the content rather than taking the content to AI, and its press material records that more than 800 firms began using the AI capabilities in 2025 and that over forty per cent of new customers select AI at purchase, which describes an option on a platform rather than the platform. Remove the models and the document management, email management, search, security and collaboration product remains and is what most of the 7,000 organisations bought. Legal AI page, Smart Answers press release of 4 March 2026 and the Software as a Service Agreement read 7 September 2026.
Citation Accuracy and Hallucination Disclosure
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
Measured accuracy is published with the test set described and the failure modes named, which no other record located in this index does at this depth. The Legal Context Engineering Benchmark, published 18 August 2026 and authored by the VP of Product, reports quality scores for the vendor's own AI answers out of 100 across three model tiers, 64.0, 67.4 and 73.9 without the Legal Context Graph against 65.1, 69.0 and 75.4 with it, alongside correct-answer counts out of 300 and the cost of a correct answer. The test set is described in full: 300 questions across ten real matters, five transactional and five litigation, assembled from public filings and court dockets falling partly outside the models' training cut-off, 874 documents and roughly sixty million characters, nothing synthetic, with six named question types including an Unanswerable class and a per-question rubric, and citations in answers checked deterministically against the corpus with human spot checks. Failure modes are named rather than buried: a worked example shows the product's own answer failing all three criteria without the context layer and still failing one with it, and the report states how a cost-per-correct-answer figure can be gamed by declining hard questions. The method is published for firms to re-run. Two limbs of the band do not bite and are named: the product answers from the firm's own repository rather than from primary legal authority, so there is no citator, and cited sources are documents the reader already holds. The report is the vendor's own internal benchmark rather than third-party validation, and the appendix PDF with per-matter scorecards was not read. Benchmark report web edition read in full 7 September 2026.
Autonomy and Oversight Model
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
Autonomy is described and oversight rests on surfaces rather than on a published control structure. The vendor states that the best AI apps run automatically in the background based on a user's actions in NetDocuments, and its MCP connectivity lets external agents orchestrate workflows across systems, which is a real claim of unattended operation. What exists on the oversight side is genuine but is infrastructure rather than a stated model: answers carry citations a lawyer can open, Exhibit B of the agreement commits to document-level activity logging reviewable for ninety days, administrators control which users and applications reach what through the Customer Service Account, and the benchmark report tells readers to read some of the answers themselves before believing any of the numbers. No located material states that the models work alongside a supervising lawyer, names the threshold at which an app acts without one, or says what happens after an output is wrong. The Our Approach to AI video on the Legal AI page was not viewed and is the rebuttal route. Legal AI page, Smart Answers press release and Exhibit B read 7 September 2026.
Operational and Outcome Evidence
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
Named customers with a figure, short of dates and a method a reader can assess. The Legal AI page carries an attributed customer statement from Jared Gullbergh of Buchanan that a single AI app was calculated to save over 1,500 hours of work a year while producing better work product, alongside attributed statements from Charly Duffy of Coghlan Duffy on the App Builder and Jeff Sabado, Director of Knowledge Management at Davis Wright Tremaine, that the tools already respect the firm's client and matter based security and ethical walls. Corporate figures are published without attribution to any deployment: more than 7,000 law firms, corporate legal departments and public sector organisations, more than 800 firms beginning to use the AI capabilities in 2025, over forty per cent of new customers selecting AI at purchase, and 150-plus integrations. What separates this from the top band is that no deployment is dated and the basis for the 1,500-hour calculation is not published, so a reader cannot assess how it was reached. Legal AI page, Ethical Walls page and Smart Answers press release read 7 September 2026.
Privilege and Confidentiality Posture
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
Substantive published commitments across most of the ground, short of the privilege limb the A band names separately. What is published and binding: Customer Repository Data is confidential under the agreement and NetDocuments disclaims ownership of it (5.2.1); use is confined to providing the Services, facilitating third-party applications the customer selects, help desk support the customer initiates, and legal demands (5.2.2); ethical walls are a documented product with access controls at user, document and workspace level, and the vendor states the AI operates inside a firm's existing permissions and walls; the ndMAX FAQ states inputs and outputs are not kept beyond producing the answer and are not used to train the model; and the position on the model provider is explicit, including an agreement with Microsoft that keeps customer content out of abuse monitoring. Two things hold it here. Privilege and work product are not addressed anywhere located, which R33 makes a required limb of the A band and not a formality. And the training statement is scoped to the LLM rather than to NetDocuments' own use, which is the narrower promise of the two. Agreement, Legal AI page, App Builder page and Ethical Walls page read 7 September 2026.
UPL and Professional Responsibility Posture
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point.
No position on the advice line was located on any surface read, for a product whose AI apps analyse NDAs, assess contract risk, produce judge analytics and answer questions about a matter. The Software as a Service Agreement was read in full and addresses use restrictions, AI restrictions binding the customer, warranties and liability without touching advice, competence or supervision; the agreement disclaims that results will meet any requirement (9.4), which is a warranty disclaimer rather than a statement about what the product is and is not. The Legal AI, App Builder and Ethical Walls pages carry no disclaimer. One limb does point the right way and is recorded rather than credited: the audience is unambiguous, since every segment page addresses law firms, corporate legal departments and public sector legal teams, and the agreement confines use to internal business purposes by named Internal and External Users, so there is no consumer-facing surface. This grade records what is establishable on the date rather than a finding that no position exists anywhere; the full text of the Terms of Use, reached only in a search excerpt, is the rebuttal route. Surfaces checked 7 September 2026.
AI Governance and Bias Disclosure
Published governance over model behaviour: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
Published substance with an owner and a testing regime, short of a governance framework and any bias disclosure. The Legal Context Engineering Benchmark is a signed technical report by the VP of Product describing what is tested and how, including a testing harness that compares the two tool surfaces before a run and refuses to proceed if they differ in any respect other than the capability under test, deterministic validation that citations map to the corpus, human subject-matter spot checks, and a published method other organisations can apply. The report also discloses its own limits: that it does not yet measure change over a matter's life, that the vendor deliberately withholds its context build cost because it varied roughly sixfold across ten matters and a single figure would mislead, and that the report was drafted with AI assistance. What is absent is the governance half of the axis. No responsible AI framework, policy or standard such as ISO 42001 was located, nobody is named as accountable for AI governance as distinct from authoring the benchmark, and nothing published addresses uneven output across matter types or populations. The unread appendix PDF carries per-matter scorecards and is the route to more. Benchmark report read 7 September 2026.
AI Safety and Data Stewardship
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
Substantive published policy covering most of the ground in a binding exhibit, short of a named subprocessor list. Exhibit B of the Software as a Service Agreement commits NetDocuments to report confirmed Unauthorized Access to the customer without unreasonable delay and within 24 hours, with root cause analysis and a mitigation plan; to encrypt Customer Repository Data in transit and at rest; to least-privilege access with removal on termination of employment; to independent penetration tests at least twice per twelve months and monthly internal and external vulnerability scanning; to document-level and administrative activity logs reviewable by the customer for ninety days with NetDocuments retaining log information for at least a year; to storage across two or more geographically separate data centres; and to disaster recovery targets of a two-hour recovery point and six-hour recovery time. Deletion is governed by the agreement at 7.4, with the customer responsible for export and NetDocuments free to destroy remaining data thirty days after the term. The gap is subprocessors: section 12.2 permits Common Subcontractors and promises Digital Notice of material changes and additions without publishing who they are, and the only named processor anywhere is Microsoft for the AI. Agreement, Exhibit B and Legal AI page read 7 September 2026.
AI Liability and Recourse
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
What the vendor stands behind is published and specific. Section 11.1 of the Software as a Service Agreement gives a defence and indemnity against third-party claims that the Services infringe intellectual property rights, with three named exclusions and modify, licence or terminate remedies at 11.3.3; section 10.3 caps each party at the annualised subscription fee paid or payable at the time of the event, with the indemnity obligations carved out of the cap; 10.1 excludes indirect and consequential loss; 10.4 preserves liability for death or personal injury caused by negligence and for fraud; 9.2 warrants that the Services will be performed with the care and skill of a professional and competent provider; and Exhibit A publishes a 99.9 per cent availability target with a graduated service credit schedule that functions as an agreed remedy. No insurance is stated. One asymmetry a buyer should read before signing, recorded here because it is what the agreement says rather than because it moves the grade: section 10.1 removes every liability limitation for the customer's breach of the AI Restrictions in section 3.5, so NetDocuments may recover lost profits, lost business opportunity, lost competitive advantage and all indirect and consequential damages without cap, while the customer's own recovery remains capped at one year of fees. Agreement read in full 7 September 2026.
Practice Systems Integration Depth
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
Real, documented integrations named across the systems legal work lives in, short of an implementer's description of what moves in which direction. The vendor states the platform integrates with more than 150 technologies including Microsoft 365, Teams and DocuSign, publishes an App Directory and partner locator, and ships ndOffice and mobile access software defined in the agreement. The 2026 additions are the substantive part: Model Context Protocol connectivity lets external AI applications and agents reach NetDocuments content and orchestrate workflows without file downloads or manual transfers, with the vendor naming ChatGPT, Claude, Claude Cowork and its legal plugin, Microsoft Copilot, Perplexity, Google Gemini Enterprise, Harvey and Legora, and stating the boundary that this operates within existing permissions, ethical walls and audit controls with no content leaving the platform. The agreement describes what a firm must do for the API: it is not included in the subscription, access is granted on request after registering an application, it may not be shared, and further terms may be imposed. What is not published on the surfaces read is per-integration depth, which lives in the unread App Directory and support documentation. Smart Answers press release, agreement section 4.4 and Legal AI pages read 7 September 2026.
Deployment Model and Data Residency
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Deployment options and residency are published and contractual, and processing is addressed separately from storage. Services Region is a defined term in the Software as a Service Agreement, and section 2.4 commits NetDocuments to store Customer Repository Data in the region specified on the Order Form and not to transfer or access it except at the customer's or a user's direction or as required by law, with sharing and out-of-region user access carved out where the customer configures it. Five regional entry points are published on the site: United States, United Kingdom, Germany, Australia and Canada, and the contracting entity itself follows the region. Tenancy is stated as the customer's private repository on the ND Network, and Customer Managed Storage Locations using ndFlexStore or ndMirror let a customer hold repository data on servers NetDocuments does not host, which is the tier difference described. The processing limb is met expressly and against the vendor's interest: section 4.2 records that help desk support may be provided by personnel or subcontractors in regions other than the customer's Service Region and requires the customer to say so in advance if its policies or law restrict that. One thing a buyer should ask about is not published: where Azure OpenAI inference runs for a customer in a given region. Agreement and site region list read 7 September 2026.
Security Certifications and Trust Center
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
Certification is real, named and contractually committed, short of evidence this channel could reach. Exhibit B of the Software as a Service Agreement commits NetDocuments to have a Type 2 SOC 2 audit for security, availability and privacy undertaken annually and to maintain ISO 27001 certification or a comparable successor, states that all data centres fall within the scope of the annual ISO 27001 audit, and records that the listed controls are verified in both. That is a named scope and a stated cadence carried in a binding instrument rather than a badge, which is more than most records on this axis have. What is missing is the accessible half: no auditor, no audit period and no report date were located, and the trust website named in the agreement, trust.netdocuments.com, is a TrustShare portal that returns a JavaScript shell with no body on this channel, so whether its Due Diligence Response material is self-serve or sales-gated could not be established. Recorded as a retrieval limit rather than as an absence, with the trust site as the rebuttal route; under the standing convention the lower tier is graded and the reason stated. Agreement Exhibit B read in full and trust site attempted 7 September 2026.
Model Supply Chain Disclosure
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
The provider is named and the architecture described; the models are not named and change notice is reserved rather than committed. The Legal AI page states that ndMAX is powered by Microsoft Cognitive Services and Azure OpenAI and their suite of OpenAI models, that only inputs and outputs are shared with the LLM, that they are not kept beyond the exchange, and that an agreement with Microsoft prevents customer content from triggering Microsoft's abuse monitoring. The same FAQ reserves the right to change what sits underneath: the solutions are described as model agnostic, so that if better models are developed they may be used, and new models are tested internally and made available as they appear. No specific model is named, no inference location is stated, and the only change-notification machinery located is the general Digital Notice of material changes to Common Subcontractors at section 12.2 of the agreement, which is not a commitment about models. One fact is deliberately not counted here: the MCP connectivity that lets a firm's own ChatGPT, Claude, Copilot or Harvey reach NetDocuments content is the customer's model choice reaching the vendor, which is extensibility and is credited on the integration axis instead. Legal AI page and agreement read 7 September 2026.
Commercial Transparency
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
The shape of the charge is published and no number is, at any tier. Every purchase path on the site ends at Contact Sales or a demo request, and no price, band or unit figure appears on any page read. What lifts this off the floor sits in the published agreement rather than on a pricing page: fees are set in an Order Form as recurring Subscription Fees with a Billing Commencement Date, section 4.2 records that a customer exceeding the End User, lawyer, document or other licence metrics stated in the Order Form is invoiced for the overage, section 4.4 caps annual increases at CPI plus five per cent no more than once a year, section 8.1 requires disputed invoices to be raised within thirty days, and Exhibit A sets out service credits as a proportion of the annualised fee. Tier names are visible in the product material, with ndMAX Enterprise identified as the tier receiving Smart Answers and expanded MCP connectivity. So a buyer can learn the unit, the escalation ceiling and the tier structure before a sales call, and cannot learn the price. No VendorPricing row is written: under the pull 7 convention a row belongs to vendors graded A or B on this axis and the grade speaks for the rest. Agreement and site read 7 September 2026.
Firm and Practice Coverage
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
Segment coverage is described with unusual precision, and the boundaries are left open. The site publishes separate pages for small, midsize and large law firms, corporations, in-house legal departments, public sector, federal government, and state, local government and higher education, plus regional pages for Asia, Australia and New Zealand, North America, South America, and the UK and Europe, and the vendor states it supports more than 7,000 law firms, corporate legal departments and public sector organisations. Practice coverage is described through the work rather than by area: the ndMAX Studio library covers automatic profiling, NDA analysis, contract risk analysis, playbook creation and application, deposition preparation, judge analytics, offer letter workflows and translation, and the benchmark corpus is split evenly between transactional and litigation matters, which is the closest thing to a statement of where the AI has been tested. Nothing published states what is not supported: no practice area, jurisdiction or language is named as out of scope, and the eDOCS line for on-premises customers is presented without a boundary against the cloud product. Site segment pages, ndMAX Studio, Legal AI page and benchmark report read 7 September 2026.
Legal Signals
What each signal meansA signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
A public policy or trust page states no training on customer content, with no matching term located in the published agreement.
A product page states no training and the published agreement carries no matching term. The ndMAX FAQ on the Legal AI page states that only inputs and outputs are shared with the LLM, that they are not kept beyond the exchange, and that they are not used to train it, and the App Builder page adds that an agreement with Microsoft ensures customer data is not used to train the Azure OpenAI model and is not subject to human or machine sensitive content monitoring. Two scope points a buyer should carry. The statement is about the LLM, which is Microsoft's model, and no located material says whether NetDocuments itself uses customer content to improve its own systems. And the Software as a Service Agreement, read in full, confines use of Customer Repository Data to providing the Services and never names training, while separately defining Platform Data as data and statistics generated in connection with use of the Services, which NetDocuments may use to analyse performance and usage in order to provide or improve the Services and may use in anonymised aggregated form for benchmarking. That clause names no model and operates on usage data rather than on repository content, so it is recorded here rather than treated as a training permission. Agreement, Legal AI page and App Builder page read 7 September 2026.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
A specific retention period is published and the customer cannot change it.
A specific retention position is published and the customer does not set it. The ndMAX FAQ states that inputs and outputs are not kept beyond the exchange of information needed to produce the output, which is a stated period of none rather than a window a customer configures, and no zero-retention setting or alternative window is offered anywhere located. Two adjacent facts that are not the same thing and are recorded so they are not read as one: an output a user saves into the repository becomes Customer Repository Data and is then governed by the customer's own retention and by the agreement's deletion terms, and Exhibit B commits to document-level activity logs the customer can review for ninety days with NetDocuments retaining log information for at least a year, which is audit metadata rather than prompt content. Legal AI page and agreement Exhibit B read 7 September 2026.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
Retrieval enforces the source system access model at query time, per user, and the vendor documents it.
Retrieval enforces the firm's access model at query time and the vendor documents it, in the unusual case where the vendor is itself the source system. Ethical walls are a documented product with access controls at user, document and workspace level and reporting on who accessed which document, and three vendor statements place the AI inside them: the Smart Answers release of 4 March 2026 states that answers are produced within the existing security, governance and ethical wall controls and that MCP connectivity operates within existing permissions, ethical walls and audit controls with no content leaving the platform, and the Legal AI page carries a named customer, the Director of Knowledge Management at Davis Wright Tremaine, saying the tools already respect the firm's client and matter based security and walls. What is not published is the enforcement mechanism at the retrieval layer itself, whether a wall is applied before or after the index is searched, and that distinction matters to a firm assessing whether an agent can infer the existence of walled content. Ethical Walls page and Smart Answers press release read 7 September 2026.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Terms commit to notice where lawfully permitted. No transparency report located.
The agreement commits to prior notice where lawfully permitted, and no transparency report was located. Section 6.2 of the Software as a Service Agreement permits disclosure of the other party's confidential information to comply with applicable law or a court order, states expressly that in NetDocuments' case such disclosure may include Customer Repository Data, and conditions it on giving prior notice of the compelled disclosure to the extent legally permissible together with reasonable assistance, at the disclosing party's expense, if that party seeks to contest the disclosure. Prior notice plus an assistance obligation is more than the bare notice most agreements in this corpus carry. No transparency report, request statistics or law enforcement guidelines page was located on any surface read. Agreement read in full 7 September 2026.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Sources are identified without stating the licence or rights basis.
The sources behind the answers are identified and no licence basis is stated for anything beyond the customer's own content. The product ships no corpus of law: Smart Answers is grounded in the firm's own document repository and matter history with citations, and the Legal Context Graph organises that same content. The rights basis for the customer's own material is allocated by the agreement rather than licensed by the vendor, since section 9.3 has the customer warrant it has the legal right to possess, store and transmit its repository data and section 5.2.1 has NetDocuments disclaim ownership of it. Two things are not established. The August 2026 extension of Legal Authorities to the Legal Context Graph implies a source of law outside the firm's repository, and no licence, provenance or update cadence for it was located. And the benchmark corpus, whose provenance is stated as public regulatory filings and court dockets, is a test set rather than a product corpus and is not credited as one. Smart Answers press release, benchmark report and agreement read 7 September 2026.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
No located public material addresses whether authority is checked for subsequent history.
No located public material addresses whether authority is checked for subsequent history. The product answers from the firm's own documents and matter history and cites back to them, and nothing read describes a treatment signal, a citator licence or a verification prompt for legal citations. The August 2026 extension of Legal Authorities to the Legal Context Graph is the one surface that might bear on this and its product page was not read. The signal's limbs bite only partly for a document management platform whose citations point at the firm's own files rather than at reported authority, and that is recorded rather than graded around. Legal AI page, Smart Answers press release and benchmark report checked 7 September 2026.
Refusal and Uncertainty Behaviour
What does the product do when the answer is not in the corpus?
The vendor documents an explicit no answer path and it is observable in the product or in published evaluation.
An explicit no-answer path is documented and observable in published evaluation. The Legal Context Engineering Benchmark of 18 August 2026 defines Unanswerable as one of six question types in its 300-question set, described as questions the record is silent on, so abstention is something the vendor deliberately tests for rather than an incidental behaviour. The report then publishes a transcript in which the agent declines to supply a date it cannot support from the documents, and grades that answer against the rubric in the open. The limit worth stating is that the published example shows abstention in the run without the context layer, and the report does not publish a per-type breakdown of how often the system abstains correctly against how often it should have; the unread appendix carries per-matter scorecards. Benchmark report read 7 September 2026.
Fabricated Citation Record
Does a public court record exist involving output from this product?
No court order, opinion or disciplinary record naming this product has been located as of the date shown. This is a statement about the public record, not a finding about the product.
No court order, opinion or disciplinary record naming NetDocuments, ndMAX or Smart Answers was located as of 7 September 2026. The AI Hallucination Cases database maintained by Damien Charlotin was searched on the company and product names alongside a general search for sanctions coverage; the records that name legal-specific products name other vendors. This is a statement about the public record, not a finding about the product. The exposure here is structurally lower than for a research tool, since the assistant answers from the firm's own documents and cites back to them rather than producing legal authority.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Public materials refer to professional responsibility in general terms without naming guidance.
Professional obligations are engaged in general terms and no AI ethics guidance is named. The Ethical Walls page frames its controls around preventing conflicts of interest, maintaining client confidentiality and ethical standards, and demonstrating compliance to clients and regulatory bodies, which is real engagement with a lawyer's duties and is already credited where it belongs, on the walls signal and on confidentiality. What this signal asks is different: whether the vendor engages publicly with the ethics opinions and regulator guidance governing lawyers' use of AI. No located material names an ABA formal opinion, a state bar opinion, a court standing order or any regulator's AI guidance, and the AI adoption resources are framed around investment and results rather than professional responsibility. Ethical Walls page, Legal AI page and resource listings checked 7 September 2026.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Public materials claim time savings without addressing billing or disclosure, and the product sits inside a fee relationship between a lawyer and a client where those savings would change the bill.
The product sits inside law firms' fee relationships with their clients and the published position on the bill is a savings claim. A named customer at Buchanan states that one AI app was calculated to save over 1,500 hours of work a year, and the Legal Context Engineering Benchmark quantifies a different saving, forty-eight per cent off the cost of a correct AI answer and roughly $940,000 a year for a 2,000-person firm, which is the firm's own AI spend rather than anything reaching a client's invoice. Nothing located addresses how AI-assisted work is recorded on a matter for billing purposes, whether time compressed by the assistant is disclosed to the client, or what a firm should tell a client whose work took an hour instead of six. The platform holds the documents rather than issuing the invoice, but it is bought by firms that bill clients for the work it accelerates, so the question applies. Legal AI page and benchmark report checked 7 September 2026.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
A current subprocessor or model provider list is published.
The model provider behind the AI is named publicly and no subprocessor list is published. A firm answering a client's AI clause can point to the ndMAX FAQ, which names Microsoft Cognitive Services and Azure OpenAI as what sits behind the generative features, states that inputs and outputs are not kept beyond the exchange and are not used to train the model, and records an agreement with Microsoft exempting customer content from abuse monitoring; the full Software as a Service Agreement and its security exhibit are published and forwardable, which covers the incident notice and control commitments a client is likely to ask about. Two limbs are short. Section 12.2 of the agreement reserves the use of Common Subcontractors with notice of material changes but publishes no list of who they are, so there is no current subprocessor register to forward. And the Data Processing Addendum, which is the artefact drafted to be passed on, is provided on request under section 6.3 rather than published. Agreement and Legal AI page read 7 September 2026.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
Some elements of the record are available, short of a document level export.
Elements of a record exist without a document-level export built for disclosure. Exhibit B of the agreement commits to logging that lets a customer review document-level events and administrative changes for the previous ninety days, with NetDocuments retaining log information for at least a year, and the Ethical Walls page describes reporting on exactly who accessed which documents; on the output side, Smart Answers returns answers with citations to the source documents. Between them a firm can establish which documents an assistant touched and what an answer was grounded in. What is not published is any per-document export covering the model used, the retrieval performed and the human verification applied, and no disclosure template or guidance for a judge's standing order on AI use was located. Agreement Exhibit B, Ethical Walls page and Smart Answers press release read 7 September 2026.