C

Counsel Stack

For the research product, the AI is the product. The vendor describes a fleet of specialized research language models built by its attorneys, engineers and legal professionals for particular practices and jurisdictions, and the evaluator independently confirms multiple specialized models built for legal research. Retrieval, synthesis and citation verification are all model driven. The company's legal marketing, lead generation and web design businesses have nothing to do with AI and are separate from the research product the benchmark measured.

Accuracy is measured independently. In the 2025 Vals Legal AI Report legal research study, Counsel Stack placed first among all participants on all three criteria, ahead of every other legal AI product, a generalist model and the lawyer baseline: 81 percent accuracy against a 71 percent lawyer baseline, 77 percent authoritativeness against 68 percent, 71 percent appropriateness against 60 percent, and a 78 percent weighted score against 69 percent.

It answered 196 of 200 questions, with four technical timeouts the evaluator published with the scores. The evaluator set the rubric and weights in advance and Am Law 100 firms supplied the questions, so the figures can be checked on the evaluator's own site. The product also ships structured citation verification, saving proof of verification automatically in the research record. The evaluator found that all systems struggled with multi jurisdiction questions and did worse than a generalist model on fifty state statutory surveys.

Autonomy is present, and oversight is implied by record keeping rather than described as a control. The product automatically saves research, conversations, sources and proof of citation verification, so a supervising lawyer can reconstruct what was done, and a published customer account describes the tool helping a small team rather than replacing people. Reconstructing work afterward is transparency, not a control point.

Searched the vendor site, deep research product page and third party coverage and found nothing on what the system does unaided, when it stops or escalates, what a lawyer must review before output is used, or what the vendor commits to when output is wrong. A traceability feature does not amount to oversight.

Testimonials stand in for deployment evidence, though they are specific about the practice. Published accounts describe a solo attorney working with one other attorney, a paralegal and an intern who replaced their practice management software and use the product for research, and a legal content creator using it for Pennsylvania and federal law. These are real descriptions of use, honest about very small firm size. Searched the vendor site, product pages and third party coverage and found no named organization with figures, a date or a method. The independent benchmark measures test performance, not production outcomes.

Searched the vendor site, the deep research product page and third party coverage, and ran targeted searches for a Counsel Stack security page, privacy policy or trust center without reaching one. Nothing is published on how client confidences are handled: no statement on training, no retention or deletion terms, no segregation model, no treatment of attorney client privilege or work product. The product saves research, conversations and sources by design, so material accumulates in it, and no published terms attach to that material.

A real position is enforced as a product gate rather than stated as a disclaimer. Access is limited to US licensed attorneys, verified at signup, and the vendor says only US licensed attorneys qualify at this time, so a nonlawyer cannot get the tool. That sets who may hold the product but not what the lawyer remains responsible for: searched the vendor site and product pages and found nothing on competence or supervision duties, the advice line, or jurisdiction limits within the United States, though the product spans federal, state and local doctrine.

Agreeing to independent measurement and to publication of the results is the governance act here. Counsel Stack was one of four systems evaluated and named in the Vals legal research study, where the evaluator and trade press both reported that major vendors declined and at least one took part but then withheld permission to publish. Allowing publication of results that included four technical failures is a real accountability position, but it is a one time evaluation, not a management system.

Searched the vendor site, product pages and third party coverage and found no AI governance framework, AI management certification, named owner of model governance, testing regime before release, or anything on uneven output across matter types, parties or populations.

Searched the vendor site, the deep research product page and third party coverage, and ran targeted searches for a security page or trust center without reaching one. No retention period, no deletion control, no access control detail, no encryption statement, no hosting disclosure, no subprocessor list and no incident or breach notification practice is published. The gap is sharpened by the product's own design: it automatically saves research, conversations, sources and timekeeping entries, so it accumulates a durable record of a lawyer's work with no published terms governing that record.

No published indemnity, liability cap, carve out, warranty on output or insurance position is published, and no customer terms of service is published as published on the property. Recorded as a pure absence on the surfaces reached. Worth naming the shape of it: this vendor's entire positioning rests on measured accuracy and structured citation verification, and it publishes nothing about who bears the loss if a verified citation is wrong.

Integration routes exist and none is documented. API access is offered on request, and a customer account describes the product replacing their practice management software, which suggests matter and time tracking built into the product rather than connected to it. Searched the vendor site, product pages and third party coverage and found no integrations page, no named connector for document management, practice management, court filing or Microsoft Word, and no API documentation available without contacting the vendor.

Searched the vendor site, deep research product page and third party coverage, and ran targeted searches for infrastructure or security documentation without finding any. Nothing was found on deployment: no tenancy statement, hosting provider, region, residency option or processing location. The product is evidently cloud delivered and reached in a browser, but that is inferred from how it is sold, not stated.

Two targeted searches for the vendor's security page, privacy policy, terms, trust center or certifications found no vendor material. A failed search is weaker evidence than a page, so this shows that a buyer doing ordinary diligence would not find security material, not that none exists, and a single link could change it. The vendor publishes an independently measured accuracy score but nothing about how it handles data.

The vendor describes its own model layer without identifying what sits underneath it. Published and independently confirmed: a fleet of multiple specialized research language models built for particular practices and jurisdictions, described by the vendor as assembled over years by its own attorneys, engineers and legal professionals. That tells a buyer the architecture is specialist rather than a thin wrapper. Searched the vendor site, the product pages and third party coverage and located no named foundation model or provider underneath that layer, no statement of where models run, no subprocessor list, and no commitment to notify customers when the supply chain changes.

Real commercial terms are published at the entry point, but no rate is. A free tier can be reached without a sales conversation: the vendor says signup needs only a name and professional email and no credit card, with free trial access to its chat product. That is a published term and a self serve route. Searched the vendor site, product pages and third party coverage and found no rate for any paid tier, no unit of charge and no tier structure, with API access by contact request.

Coverage is described in substance, and the target is stated honestly. Scope is federal, state and local US doctrine, with specialized models claimed for particular practices and jurisdictions. The vendor positions itself narrowly, for highly technical, complex, boutique, cross jurisdictional and multi doctrine research, which says where it is strong rather than claiming to cover everything. Published customer accounts come from very small firms, and no enterprise scale is claimed.

Practice areas are not listed, state level coverage is not stated, and nothing says what the product does not cover; the measured weakness on fifty state surveys comes from the evaluator, not the vendor.