Jus Mundi vs Vincent AI: how they compare in 2026

The models are the engine of a core capability layered on a product that would still function without them. Jus AI answers, assesses, compares, drafts and translates over the database, and the vendor states it has trained models on its corpus since 2019; remove the models and the search engine, filters, interactive documents, CiteMap, Wiki notes and analytics remain, which is the research database the company sold before Jus AI launched in June 2023 and still sells on the Academic and Essentials plans without the assistant. Home page, pricing page and 2023 launch release read 6 September 2026.

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 organizing 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.

Grounding is real and documented with linked primary sources, short of a testable accuracy figure. The home page states that Jus AI delivers fully cited answers from exclusive Jus Mundi data, that every reasoning step can be verified, each citation checked and each source traced to its origin, and the primary documents are the company's own database of awards, decisions, treaties and laws, kept openly accessible in a free light version so a reader can open what is cited. No accuracy figure, test set, benchmark or hallucination rate is published on the surfaces read, and the security page, which may carry more, could not be fetched. Home page and pricing page read 6 September 2026.

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 summarization 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 penalized 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.

A written commitment that the models work alongside a verifying lawyer with a real review surface, short of the full control structure. The vendor describes agentic legal reasoning that guides research with clarity and structure and exposes every reasoning step for verification, with citations traceable to source; the assistant answers and drafts for a practitioner to check. What is not published is a threshold at which any agentic step acts without review or a stated route back after a wrong answer beyond the user's verification. Home page and security-page excerpt read 6 September 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.

Testimonials and a client wall stand in for named deployments with figures on the surfaces read. The home page states that clients' experiences are the strongest signal and points to case stories about winning strategies and time saved, and a customer is quoted on the security page, unnamed in the excerpt, about confidence in uploading documents given ISO 27001 certification; the seed names Freshfields and White & Case as customers, which was not confirmed on a vendor surface read. The Meet Our Clients section and case stories were not opened and are the rebuttal route. Home page and pricing page read 6 September 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 addressed with two specific commitments and the rest of the picture could not be read. Segregation: the help center states that assistant interactions are hosted on a separate and locked database, and the vendor states confidentiality is the foundation for sensitive casework. Third-party providers: the security page names Microsoft's Azure OpenAI Service and the enterprise plan connects Jus AI to a firm's own documents. Not established: any statement on training use, since the security page returned bot detection on fetch and the user terms were not read; any retention commitment; and any treatment of privilege or work product. The subscription agreement that governs firms is separate and not published. Home page, help center excerpt and security-page excerpt read 6 September 2026.

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.

No advice line or supervision statement was located on the surfaces read. The user terms describe the service as a research platform providing know-how and resources to international law and arbitration practitioners, and the product is sold to practitioners, firms, states and academics; whether the terms disclaim legal advice or address a practitioner's duties was not established because the terms were read only in excerpt. No jurisdiction limit beyond the subject-matter scope is named. User terms excerpt and home page read 6 September 2026; the full terms are the rebuttal route.

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.

A published, independently audited governance framework, short of testing results or a named owner. The security page states that with ISO 42001 the company went further than security certification to show its AI is built responsibly and to the highest ethical standards, alongside ISO 27001 and SOC 2; the home page describes exposing every reasoning step as transparency. ISO 42001 is an audited AI management system standard and carries more weight than a self-published principles page; no test results, bias findings or accountable owner are published on the surfaces read, and the security page could not be fetched in full. Security-page excerpt and home page read 6 September 2026.

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 program, 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.

Some of the ground is covered and the surface that carries the rest could not be read. Stated: assistant interactions on a separate locked database, encryption, ISO 27001 and SOC 2 certification, Azure OpenAI as the model service. Not located on readable surfaces: a retention period, a deletion commitment, a sub-processor list or an incident-notification practice; the security page and trust center returned bot detection on fetch and are the rebuttal route, together with the privacy charter, which was not opened. Home page, help center excerpt and security-page excerpt read 6 September 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 authorized 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 center 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.

No liability position was located on the surfaces that could be read. The user terms, read in excerpt, define the services, materials and subscription structure without a located indemnity, cap, warranty or insurance position, and the subscription agreement that governs firms is stated in the Jus Connect terms to be a separate document, which is not published. This records what is locatable on the date and not a finding that no position exists; the full user terms and the subscription agreement are the rebuttal route. User terms excerpt and Jus Connect terms excerpt read 6 September 2026.

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.

Integrations are referred to without documentation an implementer could use on the surfaces read. The help center headline describes an agent-to-agent integration bringing arbitration AI to where users already work, and the AI Enterprise plan connects Jus AI to a firm's own document database; nothing read names the systems, what syncs or in which direction, and no document management system is named. Help center index and pricing page read 6 September 2026.

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 stated and neither tenancy nor region is addressed on readable surfaces. The security page names Microsoft's Azure OpenAI Service as the model service and the help center a separate locked database for assistant interactions; no region, residency option or tenancy model is stated in the material read, and the security page could not be fetched in full. Security-page excerpt and help center excerpt read 6 September 2026.

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.

Certifications are real and stated on the vendor's own surface with a trust center, short of a report reachable without asking. The security page states ISO 27001 and SOC 2 as proof of security and confidentiality by design and ISO 42001 for responsible AI, and points to a trust center for the measures and certifications; a customer is quoted on the ISO 27001 certification. No certifying body, coverage period or report route is stated in the excerpt read, and the page and trust center returned bot detection on fetch, so the access tier is not established and the lower tier is graded. Security-page excerpt read 6 September 2026.

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.

The supply chain is partly disclosed. The security page states that Azure OpenAI Service enables Jus Mundi to offer its data security and that Microsoft makes AI available to the user community under its security and privacy controls, which names the provider and hosting arrangement; the 2023 launch release describes in-house machine-learning models trained on the company's database and a GPT-powered assistant. No specific model, inference region or change-notification commitment is stated on the surfaces read. Security-page excerpt and launch release read 6 September 2026.

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 organizing 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.

A figure, a unit and a term are published. The pricing page lists individual academic access at twenty-nine euros a month excluding tax, academic institutions at two hundred and fifty euros a month on a yearly subscription, and practitioners at one hundred and twenty-nine euros a month, each limited to one subscriber with monthly or yearly billing, with law firms, chambers and states on quotation, a seven-day free trial, and the features of each tier itemized; a newer plan structure of Academic, Essentials, AI Premium and AI Enterprise appears in search, and whether the AI plans carry a published figure was not established. Pricing page read 6 September 2026.

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 coverage are described with substance and the boundary is the product's stated scope. The buyer is international law and arbitration practitioners, law firms, chambers, states, public bodies and academics; the database is stated to cover all areas of international law and arbitration on any subscription, updated daily, with institutional partnerships and publisher content such as the ICC collection as add-ons; the vendor states the scope as international law and arbitration, which is the limit. No jurisdiction or matter type within that scope is named as unsupported. Pricing page and home page read 6 September 2026.

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.