Paxton AI vs Vincent AI: how they compare in 2026

The models are the product. Research, drafting, document analysis, medical chronologies and the citator are all generative or machine learning capabilities, and there is no underlying practice system that would function without them.

The artificial intelligence is the product. Vincent is an AI assistant sold as such, running more than 20 generative workflows across research, litigation and transactions, with a hybrid pipeline combining generative and rules based components and vector search organising the underlying corpus. The vLex database it retrieves from is the grounding corpus rather than a separate workflow product that would stand without the models, which is the same structure as CoCounsel on Westlaw and is graded the same way.

Substantive disclosure with a real gap, and the strongest limb here is one the category leaders miss: the vendor built its own citator and publishes precision, recall and F1 tables for it. Published measured accuracy of 93.82 percent average on the Stanford Legal Hallucination Benchmark, broken out by task including case existence, citation retrieval and court identification, plus 94 percent on the Stanford Casehold dataset of 2,400 examples for the citator. Both are public academic datasets, so an outsider can in principle re run them, which is more checkable than a proprietary framework. Two gaps. The published results date from 2024 and no republication against the current product was located as of 29 Aug 2026, which matters under the benchmark half life rule. The AI Confidence Indicator is a confidence signal rather than an explicit no answer path.

Independent measurement, a named verification mechanism and grounding a reader can open. Vincent AI participated in the February 2025 Vals Legal AI Report, opting into six tasks, second only to Harvey, and scored between 53.6 and 72.7 percent, beating the lawyer baseline on document question answering at 72.7 percent, transcript analysis at 64.8 percent and document summarisation at 58.9 percent. Those figures sit on the evaluator's site and the vendor publishes them too. The evaluator recorded something more valuable than the scores: it praised the product for refusing to answer where it lacked sufficient data rather than returning a hallucinated answer, and noted plainly that those refusals lowered its own scores. A product penalised for declining to guess is the clearest evidence on this axis anywhere in the index. Grounding is real and mechanised: every research output carries citations to primary sources with direct links, plus a Cert Citator check confirming whether the cited law remains valid. The vendor claims randomised controlled trials showing 3.67 times more reliability than leading language models, which is a strong claim with no published method located and is recorded rather than relied on. Two limits recorded rather than deducted for. The vendor's own framing of the Vals results, meeting or surpassing human benchmarks in four of five tasks, uses a different denominator from the evaluator, which reports six tasks entered and three above baseline. And vLex agreed to participate in the later Vals legal research study then withdrew before results were published, stating the study was not designed for enterprise AI tools; that withdrawal is a real limit on how much of this product has been independently measured.

The product is user directed rather than agentic, and the vendor states plainly across the site that it provides self help services at the user's specific direction and cannot give advice, which is an oversight posture of a kind. The AI Confidence Indicator is a real review aid. What is not published is a described control structure: what the system does on its own, what a lawyer must approve, and what happens after an output is wrong. Searched the site, the platform pages, the terms of service and the help center index on 29 Aug 2026.

A real published commitment with genuine review surfaces, short of thresholds. The product is positioned as a first draft assistant, with the vendor's chief strategy officer stating publicly that lawyers still bear responsibility for validating research results even when using an AI assistant. Review surfaces are concrete: a user sees the sources an answer rests on and can exclude specific sources, every output carries citations with direct links and a Cert validity check, and the Clio Operate integration gives every interaction an audit trail with access scoped as narrowly as a single matter. The vendor frames this as giving lawyers a clear path back to the source before an answer reaches a client, negotiation, filing or courtroom. Not located as of 29 Aug 2026: what any workflow does unaided, a threshold at which it stops or escalates, and what the vendor commits to when an output is wrong.

An unlabelled customer logo strip and three attributed testimonials from named individuals at named organisations stand in for deployment evidence. No figures, no dates, no described method, and no case study located. A 2025 Legalweek Leaders in Tech Law award and a 22 million dollar Series A are recognition and funding rather than production evidence. Individual logos in the strip were not corroborated. Searched the site, the security page and the news and insights index on 29 Aug 2026.

Scale claims and testimonials stand in for deployment evidence. The vendor states Vincent is trusted by 8 of the 10 top law firms worldwide and publishes an attributed customer testimonial about the relationship, and independent benchmark participation is real. But a share of top firms is a market claim without names, and the Vals result is test set performance rather than a production outcome. Searched the vendor site, the Vincent product pages, the Clio press material and the vLex news section on 29 Aug 2026 and located no named customer paired with figures, a date and an assessable method.

Confidentiality is asserted in general terms and the documentation behind it points the other way, so this sits at C on the claim rather than lower. The home page describes a secure closed model prioritising the privacy of legal inquiries. Reading the published documents on 29 Aug 2026: the sitewide footer states that communications with the vendor are not protected by attorney client privilege or as work product; the terms of service require a user to warrant that submissions do not constitute confidential information, in a section scoped to questions, comments, feedback and other information about the Services rather than to uploaded documents; the privacy policy is scoped to website and account information, lists no document or prompt data, and states that the vendor does not process sensitive information, on a product that ingests medical records. No training prohibition, no retention period for uploaded documents, and no matter level segregation was located in any of them.

Substantive published commitments with the best retention answer on the index, held off an A by one limb. Retention is published with a period and customer control: the customer selects the retention window for conversation logs and files, defaulting to one year, after which data is permanently deleted. Segregation is documented at matter level through the Clio Operate integration, which the vendor states inherits the firm's existing permissions, ethical walls and audit trail with access scoped as narrowly as a single matter, and Vincent itself is described as multi tenant with each client's data logically isolated. Role based access control governs internal access. Zero retention agreements are stated with the language model providers, and Clio states no sensitive data is retained by the model. What holds this off an A is the training limb, which the A band names alongside retention and segregation: searched the vendor site, the Vincent pages, the security knowledge base article and the Clio press material on 29 Aug 2026 and located no statement of whether vLex itself may train on customer content. Provider retention and vendor training are different questions and the located material answers only the first.

A real published position on the advice line, more specific than most of this market, carried in a footer on every page: not a law firm, not a substitute for an attorney, communications not protected by attorney client privilege or as work product, and no advice, explanation, opinion or recommendation about legal rights, remedies, defenses, options, selection of forms or strategies. Jurisdiction is addressed in part, since the terms state the service is not intended for use where that would be contrary to local law. Two things keep it off an A. Nothing published addresses a lawyer's competence and supervision duties. The self help framing sits oddly against marketing aimed at attorneys, so who the disclaimer is written for is not entirely clear.

A real position on the lawyer's continuing responsibility, published in substance rather than as a disclaimer. The vendor positions Vincent as a first draft assistant and its chief strategy officer has stated publicly that lawyers still bear responsibility for validating their research results even when using an AI assistant, with the Cert validity check framed as giving a clear path back to the source before an answer reaches a client, negotiation, filing or courtroom. The audience is professional throughout, spanning law firms from solo through global, corporate legal departments, courts, government agencies, bar associations and academia, with no consumer surface located. Short of an A because competence and supervision duties are not addressed as such, and because no jurisdiction limits are stated despite coverage across more than 100 countries where professional rules differ materially. Recorded with the sourcing noted: the responsibility statement appears in trade coverage quoting an executive rather than in a standing published policy.

Searched the home page, platform pages, security page, the open trust center, the terms of service, the privacy policy and the news and insights index on 29 Aug 2026. No governance position for model behaviour was located: no named internal owner, no pre release testing regime, no responsible AI statement, and nothing on uneven output across matter types, parties or populations. The trust center lists security control families, which is a different subject. The white papers index was not reviewed item by item, so this is rebuttable with a single link.

Principles and evaluation are published without a governance mechanism a buyer could audit. Real elements exist: participation in an independent benchmark with results published including unflattering ones, a stated hybrid architecture combining generative and rules based components for reliability, a claimed randomised controlled trial programme, and Vincent Studio described as embedding firm processes into workflows with consistency and governance. Searched the vendor site, the Vincent pages, the security knowledge base and the Clio press material on 29 Aug 2026 and located no published AI governance framework, no AI management certification such as ISO 42001, no named owner of model governance, no pre release testing gate, and nothing on uneven output across matter types, parties or populations.

A generic privacy policy covers the website without addressing what happens to documents and prompts after processing. The security page does publish real controls: encryption in transit and at rest, quarterly access reviews on a least access principle, vendor risk management and a security development lifecycle, and the open trust center lists control families. But the privacy policy is scoped to account and marketing data, its retention section speaks only to the life of an account, and the terms say only that certain transmitted data is maintained to manage performance of the service. No retention period for uploaded documents, no subprocessor list, and no incident or breach notification practice was located as of 29 Aug 2026.

Substantive published policy covering most of the ground, at a level of specificity few records here reach. Published in the vendor's security knowledge base: customer selectable data retention for conversation logs and files defaulting to one year with permanent deletion after, role based access control restricting internal access to authorised personnel under documented circumstances, logical isolation per client in a multi tenant application, encryption described, and regular independent assessments with real time monitoring to identify vulnerabilities. A trust centre is published for obtaining certification reports. Zero retention agreements bind the language model providers. Not located as of 29 Aug 2026: a named subprocessor list, and an incident or breach notification practice.

Liability is addressed only through a standard limitation clause that disclaims the exposure the product creates. The published terms disclaim all warranties including any representation about the accuracy or completeness of content, cap total liability at the amount paid in the six months before the claim, require the user to indemnify the vendor, and state that the vendor has no liability for loss or corruption of transmitted data. No indemnity running to the customer, no warranty on output, and no insurance position was located. This is a C rather than a D because the allocation of loss is published and a buyer can read it before signing.

Searched the vendor site navigation, the Vincent product pages, the security knowledge base, the vLex news section and the Clio press material on 29 Aug 2026. No published indemnity, liability cap, carve out, warranty on output or insurance position was located, and no customer terms of service or master agreement was located as published on the property. Recorded as a pure absence on the surfaces reached. Rebuttable with one link, and worth noting that for a vendor of this scale, now part of a larger public facing company, published terms may well exist somewhere the searches run did not surface.

Searched the home page, the platform navigation, the pricing page, the security page and the help center index on 29 Aug 2026. No integration into document management, matter management, contract lifecycle or filing systems was located in vendor material, and the platform navigation lists none. A third party review reports a Microsoft Word add in, which is not vendor material and does not move this axis. The product is positioned as a browser based tool for solo and small firm users, which is consistent with standing alone.

Real integrations exist, are named, and include the legal specific ones that matter. Named: Microsoft Word, Microsoft Outlook, and iManage, the document management system a firm facing research product most needs, plus a deep integration with Clio Operate under which Vincent runs inside the firm's existing permissions, ethical walls and audit trail. That last is integration at the permission layer rather than the file layer, which is materially deeper than a connector. Vincent Studio adds a no code workflow builder for enterprise customers to embed firm processes. Not located as of 29 Aug 2026: per integration documentation describing what each connector moves, in which direction, and what an administrator configures, and no integrations index page was reached.

Cloud delivery is implied and the tenancy model is not stated. The terms of service do state that the service is hosted in the United States and that a user accessing it from elsewhere consents to transfer and processing there, which is a real residency statement and is why this is not lower. Not located as of 29 Aug 2026: any tenancy model, any regional option, any single tenant or private deployment, and any statement of where processing happens as distinct from where data is stored.

The tenancy model is stated and residency is not. The vendor publishes that Vincent is a multi tenant application with each client's data logically isolated and secured, which is a clear and honest statement of the deployment model and is more than several records here disclose. Searched the vendor site, the Vincent pages, the security knowledge base and the Clio press material on 29 Aug 2026 and located no hosting provider, no named regions, no customer selectable residency, and no statement of where processing happens as distinct from where data is stored. For a product serving more than 100 countries and marketed on cross border capability, published residency options would be expected and none was located.

Badges appear with no scope, no date and no report available. There is an open trust center, which is full credit as an access route: it lists SOC 2, ISO 27001 and HIPAA as compliant, GDPR as in progress, and names security control families. What it does not carry is an attestation report, a coverage period, an audit scope or the name of an auditor. Three noun problems compound this and were checked directly on 29 Aug 2026. The security page says the vendor rigorously adheres to these standards, which is adherence rather than attestation. The pricing page describes the platform as HIPAA certified, and HIPAA has no certification scheme, and its annual plan card cites ISO 20701, which is not an existing standard, while the monthly card on the same page cites ISO 27001. Separately, section 1 of the terms of service states that the services are not tailored to comply with industry specific regulations including HIPAA and that a user whose interactions are subject to such laws may not use them, which is what the agreement says alongside a marketing claim of HIPAA compliance and a product that sells medical chronology and medical billing features.

Certification is named, dated and reachable, which is the combination this axis rewards. SOC 2 certification was obtained in January 2025 and ISO 27001 certification achieved in June 2025, both stated with the month and year rather than as undated badges. A vLex Trust Center is published as the route to obtain the official certification reports for a firm's due diligence, alongside an account representative path, which under the three tier test is a self serve request flow rather than a sales gate. The vendor further states its security systems undergo regular independent assessments and real time monitoring. Short only of naming the auditing firm and stating the audit coverage period, neither of which was located as of 29 Aug 2026.

Searched the home page, platform pages, security page, the open trust center, the terms of service and the privacy policy on 29 Aug 2026. No model provider is named, no architecture is described, no subprocessor list exists, and there is no commitment to notify customers when the supply chain changes. The vendor refers to operating within a secure closed model without identifying what sits underneath.

The vendor describes its architecture and the terms binding its providers without naming them. Published: a hybrid pipeline combining generative and rules based AI, vector search organising more than one billion documents, and zero retention agreements with large language model providers. That tells a buyer the commercial terms that govern the model layer, which is the part most vendors omit. What it does not tell them is who is in it. Searched the vendor site, the Vincent pages, the security knowledge base and the Clio press material on 29 Aug 2026 and located no named model provider, no statement of where models run, no subprocessor list, and no commitment to notify customers when the supply chain changes.

Real pricing is published for part of the range with the enterprise tier withheld, which is the B band exactly and two letters above anything else on this index so far. The pricing page publishes a figure, a unit and a term: $499 per user per month, or $2,999 per user per year billed annually, with the plan contents itemised. Signup is self serve with a 7 day free trial, and the terms of service disclose the exact card authorization hold of $499. The enterprise tier is custom and volume based with no figure, no floor and no seat minimum published, and what implementation adds is not stated for either tier, which is what holds this off an A.

Checked the vendor site navigation, the Vincent product pages, the vLex home page and the Clio press material on 29 Aug 2026. No pricing page was located, no rate is published, no unit of charge is stated and no tier structure appears on the surfaces reached. Every commercial path located terminates in a demo or contact request. No third party pricing figure was located either, beyond a general observation in review material that pricing may be competitive against the two largest incumbents, which is not a figure.

Segment and practice coverage is described with substance. Named segments are solo attorneys, small and midsize firms and in house teams, with dedicated pages for personal injury, family law, employment law, criminal law and corporate law. The research coverage page is an unusually concrete boundary statement in one dimension, naming every federal and state database included by jurisdiction and court, which tells a buyer exactly where coverage stops. Not located as of 29 Aug 2026: any treatment of government or court use, and any statement of which firm sizes or practice areas the product is not built for.

The most complete coverage disclosure on the index. Jurisdictional scope is quantified and specific: more than one billion legal documents across more than 100 countries, United States primary law from Fastcase, and more than 850 million court records underpinning judge, opposing counsel and party profiling. A dedicated Compare Jurisdictions workflow and a 50 State Survey workflow exist precisely because multi jurisdictional coverage is the product's stated strength, and the evaluator independently recorded that support for international matters is a significant strength that may offer utility unmatched by other tools for global firms. Functional coverage is enumerated at more than 20 prebuilt workflows spanning research, litigation, transactions and litigation intelligence. Segment coverage runs from solo and small firms through the Fastcase heritage to global firms, corporate legal departments, courts, government agencies, bar associations and academia. The published limits are real too: the evaluator noted the study covered only a small slice of US capability, and the vendor itself stated that document summary tasks do not draw on its legal database, which is a vendor volunteering where its distinctive asset does not apply.

Read the terms of service dated 27 Feb 2026, the privacy policy dated 20 Feb 2026 and the security page on 29 Aug 2026. No located term or policy addresses training on customer content either way. The contribution license clause grants the vendor rights to access, store, process and use information the user provides, following the privacy policy, and the privacy policy is scoped to website and account information and does not mention uploaded documents, prompts or model training. Home page language about a secure closed model is not a training commitment.

The model provider layer is addressed and the vendor's own position is not. Published: zero retention agreements with the language model providers, and a Clio statement that no sensitive data is retained by the model. Those are retention commitments binding third parties, and retention and training are different questions. Searched the vendor site, the Vincent product pages, the security knowledge base article and the Clio press material on 29 Aug 2026 and located no statement of whether vLex itself may use customer content to train or improve its own systems, either way. Recorded as silent under the rule that a value is never inferred from the absence of a contradiction, and specifically not inferred from a provider side retention term.

The terms of service acknowledge that data transmitted to the service is maintained for the purpose of managing performance, with no period stated and no customer control described. The privacy policy retention section speaks only to personal information for the life of an account and does not address documents or prompts. No zero retention option was located as of 29 Aug 2026.

REGRADED 29 Aug 2026 after the value set was amended; previously recorded at disclosed fixed with the mismatch flagged. The best retention disclosure located anywhere on this index. Published: the customer selects the retention period for conversation logs and files, the default is one year, and after the selected period the data is permanently deleted. That answers all three parts of the buying question this signal asks, whether the window is disclosed, whether the customer controls it, and what happens at the end. Now recorded at customer controlled, which is the value added specifically because this record did not fit the original four. Short of the top value only because no material states that zero retention is a selectable setting, which is a real distinction for a firm that wants nothing persisted at all rather than persisted briefly.

Searched the home page, platform pages, security page, the open trust center, the terms of service and the help center index on 29 Aug 2026. No vendor material addresses ethical walls or matter level segregation. No document management system integration was located either, so there is no source system permission model for retrieval to inherit. The security page documents least access controls for vendor personnel, which is a different question.

First positive value on this signal in the index. In the Clio Operate integration the vendor states Vincent works within the permissions, ethical walls and audit trail already established in that system, with access scopable as narrowly as a single matter or as broadly as the firm and every interaction visible. That is retrieval operating inside the firm's existing access model rather than a parallel model the firm must keep aligned, which is precisely what this signal's positive value describes, and ethical walls are named as such rather than implied. Two limits recorded rather than deducted for. The inheritance is documented for the Clio Operate integration specifically, and no equivalent statement was located for Vincent used standalone or through the iManage connector, where the product's own multi tenant logical isolation and role based access control apply. Per user enforcement at query time is implied by scoping to a single matter but is not stated in those words.

Read the terms of service and the privacy policy in full on 29 Aug 2026. Neither contains a government or law enforcement request clause. The privacy policy section on when and with whom information is shared names only business transfers such as a merger or acquisition. No commitment to notify a customer of a third party request was located, and no transparency report exists.

Searched the vendor site, the Vincent product pages, the security knowledge base article and the Clio press material on 29 Aug 2026, and no published customer agreement or data processing agreement was reached. No clause committing to notify a customer of a government or law enforcement request for their data was located, and no transparency report was located. Noted for a reader: the published retention control means a firm can limit how much history exists to be requested, which reduces the exposure without addressing the notification question.

The vendor publishes a research coverage page that names its source databases individually rather than by jurisdiction alone: United States Code, the Constitution, the Code of Federal Regulations, every federal circuit and the Supreme Court, six named administrative tribunals, and for each state and territory the statutes, administrative code, court rules and named courts. That is unusually specific for this market. What is not stated is the licence or public domain basis for any of it, and no update cadence or lag is published.

Sources are named and the rights basis is stated, which is the positive value. The corpus is identified as vLex's own global legal database combined with United States primary law from Fastcase, both now owned by the same company, and the rights basis is stated directly rather than left to inference: the legal authority underlying Vincent's research is owned rather than licensed from a third party. Scale is quantified at more than one billion documents across more than 100 countries plus more than 850 million court records. Ownership of the corpus is the strongest possible answer to the question this signal asks, because it removes the licence dependency entirely. Short of the full value only on cadence: searched the vendor site, the Vincent pages and the Clio press material on 29 Aug 2026 and located no update frequency or lag for any jurisdiction, and no per jurisdiction completeness statement.

The vendor built and surfaces its own treatment signal, the Paxton AI Citator, described as patent pending, which reports whether a case has been overruled, affirmed, questioned or cited by later cases and returns explanations in Bluebook format. The method is described and quantified: evaluated against the Stanford Casehold dataset of 2,400 examples with published precision, recall and F1 figures by case type. It is the vendor's own signal rather than a licensed commercial citator such as Shepard's or KeyCite. The published evaluation dates from 2024 and no republication against the current product was located.

The first product on this index to answer the good law question properly. Every research output carries a Cert Citator check confirming whether the cited law remains valid, surfaced with the authority rather than as a separate step, and Cert is a named citator with a traceable origin: it derives from Judicata software acquired by Fastcase before the vLex merger and is now owned by the same company. Recorded at licensed citator as the closest fit, with one qualification stated rather than hidden: that value describes treatment signals from a named commercial citator, and here the citator is named and commercial but owned by the vendor rather than licensed in, which is a stronger position for a buyer since it cannot be withdrawn by a third party. Coverage limit recorded: Cert is a United States citator and no equivalent treatment mechanism was located for the international corpus, so the good law check does not extend across the 100 plus countries the research product covers.

The product exposes an AI Confidence Indicator, announced alongside the vendor's benchmark results and described as helping a lawyer judge the reliability of a response. Searched the site, platform pages, benchmark posts and help center index on 29 Aug 2026 and located no explicit no answer or abstention path, and no published statement of what the product does when it cannot ground an answer.

Abstention is independently observed and it cost the vendor measured points, which is the strongest evidence on this signal in the index. The February 2025 Vals evaluator explicitly praised Vincent AI for refusing to answer questions where it did not have sufficient data rather than giving a hallucinated answer, and recorded that those refusals negatively affected its scores. A product that accepts a lower benchmark result rather than guess is demonstrating the behaviour this signal exists to find, and the evidence comes from an outside evaluator with nothing to gain from the finding. Recorded at documented rather than the top value because the top value requires the vendor to document an explicit no answer path and that documentation was not located: searched the vendor site, the Vincent product pages and the vLex news section on 29 Aug 2026 and found the behaviour observed by others rather than described by the vendor. A reader should weigh that the behaviour is evidenced but not promised.

No court order, opinion or disciplinary record naming this product has been located as of 29 Aug 2026. Instrument searched: the AI Hallucination Cases database maintained by Damien Charlotin, which tracks court decisions worldwide involving AI generated hallucinated content and records the AI tool implicated where it is known. Also checked published 2026 sanctions summaries and secondary sanctions trackers. The entries located name filers, and in some rows other products, rather than this one. This is a statement about the public record on the date shown and not a clearance, and it is bounded by what that database covers.

No court order, opinion or disciplinary record naming this product has been located as of 29 Aug 2026. Instrument searched: the AI Hallucination Cases database maintained by Damien Charlotin, which tracks court decisions worldwide involving AI generated hallucinated content and records the AI tool implicated where it is known. Also checked published 2026 sanctions summaries and secondary sanctions trackers. The entries located name filers, and in some rows other products, rather than this one. This is a statement about the public record on the date shown and not a clearance. Noted for context: an earlier record on this index recorded a fabricated citation matter in which a Casetext or vLex product was named by a filer as a possible source; that entry was recorded as filer conduct rather than a product finding and is not restated here as a finding against this vendor.

Searched the home page, platform and practice area pages, the news and insights index, the white papers index and the webinars page on 29 Aug 2026. No engagement with any named ethics opinion was located, including ABA Formal Opinion 512 and state bar guidance. The sitewide footer carries a specific unauthorized practice disclaimer, which is a statement about the vendor's own role rather than engagement with the guidance its buyers are bound by.

Searched the vendor site, the Vincent product pages, the vLex news section and the Clio press material on 29 Aug 2026. No engagement with any named ethics opinion or professional guidance was located, including ABA Formal Opinion 512 and state bar guidance. Worth recording as context rather than credit: the company has a substantial bar association presence inherited through Fastcase, which supplied research tools to state bar memberships, so a relationship with bar associations exists without engagement with their AI guidance being published.

Vendor material is framed around time saved, describing streamlined document analysis to save time and accelerate decisions, with testimonials quoting time savings. Searched the site, platform pages, help center index and legal pages on 29 Aug 2026 and located no per matter record of AI assisted work intended for fee purposes, and no published guidance on billing, fee or client disclosure treatment.

Searched the vendor site, the Vincent product pages, the vLex news section and the Clio press material on 29 Aug 2026. No per matter record of AI assisted work intended for fee purposes was located, and no guidance on billing, fee or client disclosure treatment was located. Time savings are referenced in general terms, such as workflows turning hours of research into minutes, without a quantified savings claim tied to a matter. Recorded as not addressed rather than at the savings claims value because the located framing is capability description rather than a savings figure. Noted for a reader: the Clio Operate integration records every interaction in an audit trail, which is the raw material for such a record without being published as one.

The open trust center publishes compliance badges and security control families with no gate, which is a real access route. What a firm would need to satisfy a client AI clause is absent: searched the trust center, the security page, the terms of service and the privacy policy on 29 Aug 2026 and located no subprocessor list, no statement of which model providers see customer content, and no client facing consent or notification material. There is also no data processing agreement published.

A firm can assemble much of what a client AI clause asks for without a bespoke negotiation, though not all of it. Available without an agreement in place: a published trust centre as the route to official SOC 2 and ISO 27001 certification reports for a firm's due diligence, dated certifications, a published retention policy with customer control and permanent deletion, a stated position that language model providers are bound by zero retention agreements, and a statement that the legal authority behind the product is owned rather than licensed. Recorded at on request because the certification reports themselves sit behind a trust centre request or an account representative rather than being published, and because the artifacts this signal specifically names were not located: no subprocessor list, no naming of which model providers see client content, and no client facing consent or notification pack.

Some elements of a disclosure record are available. The product returns citations with the relevant sources highlighted, and the citator produces a treatment result with a written explanation, so sources retrieved and their status can be shown per answer. Two elements are missing: the model used is not disclosed anywhere on the property, and no per document export covering model, sources and human verification together was located as of 29 Aug 2026.

More of a disclosure record exists here than for most, assembled from several published features rather than offered as one. Through the Clio Operate integration every interaction is visible in an audit trail scoped to the matter. Every research output carries citations to primary sources with direct links plus a Cert validity check, so what was relied on and whether it was still good law are both recorded in the work product. A user can see and exclude the sources an answer rests on, which leaves a trace of human intervention. Two elements are missing: no single per document export covering model used, sources retrieved and human verification together was located, and no model is named anywhere in published material so the model used could not be stated. Recorded at partial record on that basis.