CoCounsel Legal vs Harvey: how they compare in 2026

The artificial intelligence is the product. CoCounsel Legal exists only as a generative and agentic layer, rebuilt in 2026 on an agent framework so it plans, reasons and executes multi step legal work. Remove the models and there is no product, only the underlying Westlaw and Practical Law content, which the parent sells separately as its own products.

The models are the product. Assistant, Vault, Knowledge and Workflow Agents are all generative systems, and there is no underlying document or workflow system that would stand without them. Vendor material describes every module in model terms.

CORRECTED 29 Aug 2026, second correction to this row. Previously graded B on the finding that no accuracy measurement was published, and carrying a remark that the largest company on the index published less measurement than far smaller vendors. That finding was wrong and the remark was the opposite of the truth. It came from searching only the vendor's own surfaces and not the independent evaluator's. Grounding was already strong and is unchanged: the retrieval method is described rather than asserted, with the research agent identifying key legal concepts, selecting Westlaw tools including KeyCite, the Key Number System and annotated statutes, running multiple search cycles and synthesising a structured result. Output grounds to primary authority the reader can open, with inline citations, a Sources tab carrying direct links and highlighted excerpts, and KeyCite validity flags shown with the authority. A citation level verification feature, Deep Research Verify, is published, prompting guardrails restrict output to human validated content rather than the open internet, and all of it sits under a named published framework. What is now added is measurement, and it is independent. CoCounsel participated in the February 2025 Vals Legal AI Report, a third party benchmark against a measured lawyer baseline, and scored between 73.2 and 89.6 percent across four tasks, each exceeding the lawyer baseline by more than 10 points, with a 79.5 percent average that was the highest of any participating vendor and a top score of 77.2 percent on document summarisation. Those figures sit on the evaluator's own site and the vendor also publishes them on its legal blog, so a buyer can reach them from either direction. The vendor separately publishes CoCounsel Bench, its own evaluation framework, under which experienced attorneys reviewed performance across 50 complex tasks each estimated at six lawyer hours, reporting that the product produced a stronger response than the expert written reference answer on nearly 40 percent of them. Two limits recorded rather than deducted for: the February 2025 study measured task accuracy such as document question answering and summarisation rather than citation validity or hallucination rate specifically, and this vendor did not participate in the later Vals study that measured citation authoritativeness.

CORRECTED 29 Aug 2026, second correction to this row. Previously graded B on two stated grounds, both of which are now resolved. Ground one was that BigLaw Bench is the vendor's own framework rather than independent evidence. That remains true of BigLaw Bench, but independent evidence also exists and was missed: Harvey Assistant participated in the February 2025 Vals Legal AI Report, a third party benchmark against a measured lawyer baseline, and was evaluated across six tasks scoring between 65.0 and 94.8 percent, surpassing the lawyer baseline on five of the six, with 94.8 percent on document question answering at 24.7 points above baseline and 77.8 percent on scanned and messily formatted court transcripts at 24.1 points above baseline, all at sub minute response times. Those figures sit on the evaluator's own site and are checkable without reference to any vendor claim. Ground two was that the citator and refusal limbs failed. That was a double count and is withdrawn: both are separately measured by their own signal rows on this record, and applying them again to the capability grade penalised the same absence twice. It was also inconsistent, since three legal research vendors on this index hold an A on this axis with both of those signals recorded as not addressed. The vendor's own disclosure is unchanged and remains substantial: BigLaw Bench with task categories and grading rubrics on a public repository, measured hallucination rates and source scores by model, a hallucination defined as a factual claim disprovable against a source of truth with reasoning errors tracked separately, and output linking to the specific document passages supporting each assertion. Two limits recorded rather than deducted for: the February 2025 study measured task accuracy rather than citation validity or hallucination rate specifically, and this vendor did not participate in the later Vals study that measured citation authoritativeness. The full BigLaw Bench dataset also sits behind a direct request rather than open publication.

A real published commitment with documented review surfaces, short of the full control structure. The vendor states that its AI is built with human oversight at its core, publishes a Fiduciary-Grade AI framework built around transparency and accountability, and the product is described as showing its plan before executing: it interprets a request, builds a matter specific plan, and returns work product with traceable citations the reader can open and check. Not located as of 29 Aug 2026: the threshold at which an agent stops and returns to a lawyer, and what the vendor commits to when an agent is wrong. The August 2026 shift from structured skills to a fully agentic experience makes the missing threshold more material than it would have been a year ago.

States in published material that the product is designed to assist lawyers rather than replace them and that it is built to make verification easy. Review surfaces are real and documented: inline links to source passages, role based permissions and conditionals in the workflow builder, and admin level workspace governance. Not located as of 29 Aug 2026: the threshold at which an agent stops and hands back to a lawyer, or what the vendor commits to when an agent is wrong.

Named customers, dated, with figures and an assessable method. The vendor publishes individual case studies naming Justly Prudent, reporting five times return on investment and a doubling of litigation capacity, and Brinks on reducing outside counsel spend, alongside a commissioned Forrester Total Economic Impact study dated 2026, which is a published methodology an outside reader can assess and whose commissioned status is disclosed on the page. Also publishes aggregate figures with the population stated, including 2,700 corporate customers using the CoCounsel family and a 61 percent reduction in time to draft correspondence. Deeper than anything else on the index so far on this axis.

Named customers appear in vendor material with attributed quotes, including Blank Rome on the iManage integration and a published Burges Salmon selection story. Vendor states 700 plus customers across 58 plus countries. Not located as of 29 Aug 2026: dated outcome figures with a stated method a reader could assess, which is what separates this from an A.

Substantive published commitments, and unusually the answer is given directly in a published FAQ rather than implied. The vendor states plainly that customer content and prompts are not used to train or improve the product or the underlying models, that third party model partners including OpenAI and Google are contractually prohibited from training on customer data, and that systemic controls turn off those third parties' abuse monitoring so there is no human access to customer data. Also publishes a zero retention API architecture and data residency controls. Two gaps hold this off an A. Attorney client privilege and work product handling is not addressed directly in located material. Matter level segregation is not documented on the vendor's own pages, although the product is matter centric by design.

Substantive published commitments: no training on customer data by default, a contractual prohibition on model providers training, zero data retention enforced on model providers, logical workspace separation, role based access, ethical wall sync with the firm's own walls provider, and processing in the EU, Switzerland or Australia. Two gaps keep this off an A. The security page defines customer data as uploaded documents and customer content as queries and responses as separate contractual terms, so the no training commitment reads plainly on one and not on both. Attorney client privilege and work product handling is not addressed directly in located public material as of 29 Aug 2026.

The claim is made and the documentation is not published. The product is sold squarely to licensed professionals, is distributed to more than 200 law schools and 120,000 law students, and the vendor's framing is that the system works at the level of a senior associate under professional supervision. Searched the product page, the AI hub, the trust center and the legal blog on 29 Aug 2026 and located no published position on the advice line, no treatment of competence and supervision duties, and no statement of jurisdiction limits. The senior associate framing is a capability claim rather than a professional responsibility position, and for a product marketed to students it is the place a position would matter most.

One sentence in a security blog post states the product is designed to assist lawyers rather than replace them. Checked the vendor site, security page, security addendum and help center on 29 Aug 2026 and did not locate a published position on the advice line, on competence and supervision duties, or on jurisdiction limits. The intended audience is unambiguously lawyers and legal departments, which is why this sits at C rather than lower.

A published governance framework with real substance, short of testing results or a named owner. The vendor publishes a named standard it calls Fiduciary-Grade AI with a buying framework attached, describes a dedicated research organisation of more than 1,000 specialists, states a figure for annual responsible AI investment, and answers a direct published question on bias: prompting guardrails restrict output to a clearly defined and limited set of human validated content rather than the open internet. That is a stated mechanism, not a principles page. Not located as of 29 Aug 2026: a named internal owner of model governance, published pre release testing results, or any disclosure of uneven output across matter types, parties or populations. The bias answer describes a content restriction rather than testing for disparate output.

CORRECTED 29 Aug 2026, third correction to this record. Previously graded C on the finding that no AI specific governance regime was located and that the published testing was security testing rather than model behaviour testing. That was wrong, and it came from reading the vendor's marketing surfaces rather than opening its trust centre, where the governance material actually sits. What is published on the trust centre, publicly and without a request: ISO/IEC 42001:2023 certification, the international standard for AI management systems, accompanied by a published Statement of Applicability, which is the document identifying which controls apply and why and is therefore a published scope rather than a bare badge. Alongside it, AIUC-1 certification, an AI specific assurance standard, conducted by Schellman, which the vendor states is the first accredited AIUC-1 certification body, and which the vendor describes as validating adversarial testing and its AI security programme specifically. EU AI Act conformity is separately listed. The trust centre carries a dedicated AI section with AI Governance, AI Monitoring and AI Overview items, and an AI Acceptable Use Policy sits in the published policy set. ISO 27701 for privacy information management and an IRAP attestation are also held. Two independent AI specific certifications, one of them adversarially tested, with published statements of applicability and a named accredited certifier, is the strongest AI governance position on this index, ahead of the four other A grades on this axis, each of which rests on ISO 42001 alone or on a single certification plus a framework document. One gap remains and is recorded rather than waived: no disclosure was located about uneven output across matter types, parties or populations, so bias specifically is still unaddressed, and no named individual owner of model governance was located.

Substantive published policy covering most of the ground. SOC 2 Type II and ISO 27001 maintained, ISO 42001 certification referenced for the AI management system, a zero retention API architecture published, data residency controls, 24 hour security monitoring and threat detection, corporate identity management integration, and a corporate trust center carrying the information security framework. Not located as of 29 Aug 2026: a current named subprocessor list for this product, and a stated incident and breach notification practice. The two named model partners are disclosed in the AI FAQ rather than in a maintained subprocessor list, which is the remaining element of the A bar.

Publishes retention under customer control with documented vault retention triggers and deletion timelines, role based access control, logical workspace separation, encryption in transit and at rest, and a current named subprocessor list with an update FAQ. Not located as of 29 Aug 2026: a published incident and breach notification practice, which is the remaining element of the A bar.

Searched the product page, the AI hub, the trust center, the corporate terms of use and the segment plans pages on 29 Aug 2026. No published indemnity scope, liability cap, carve out, warranty on output or insurance position was located for this product. Commercial terms are reached through a negotiated subscription agreement. The Fiduciary-Grade AI framework speaks to accountability as a design principle rather than to who bears the loss when output is wrong, and a fiduciary framing without a published liability position is exactly the shape this axis exists to record.

Checked the vendor site, its published legal pages including the security addendum and the subprocessor FAQ, and the trust center on 29 Aug 2026. No published indemnity scope, liability cap, carve out, warranty or insurance position located. Commercial terms appear to be reached through a negotiated enterprise agreement rather than published.

Real and documented integrations, short of implementer level depth. Named connections span document management with iManage, NetDocuments and SharePoint, Microsoft 365, the parent's own HighQ collaboration platform and Legal Tracker, contract lifecycle through Icertis, plus DeepJudge, Supio and Smokeball, and a Word add in carrying a personal clause library. An MCP integration lets a lawyer reach the product from a separate AI client without switching platforms, which is a genuinely forward integration surface. Not located as of 29 Aug 2026: documentation describing what each integration moves, in which direction, and what an administrator must configure.

Documented native integrations with iManage, NetDocuments, SharePoint and OneDrive, Google Drive, Box, Microsoft Word, Outlook, EDGAR and PitchBook, plus an MCP connector library. Help center articles describe what each integration moves, in which direction, what an admin must configure, and what a given integration does not support. The iManage connection is a direct OAuth integration with an embedded web extension rather than third party middleware.

Deployment is stated clearly with partial residency detail. The vendor publishes data residency controls that keep information within required geographic boundaries, operates a rollout sequenced by country across the United States, Canada, the United Kingdom and Australia, and holds FedRAMP In Process status for the underlying Westlaw and Practical Law products with stated investment toward authorization for this one. Not located as of 29 Aug 2026: the list of available regions, the tenancy model, what changes between tiers, and where processing happens as distinct from where data is stored.

States processing in the EU and Switzerland or Australia for customers with data localization requirements, and states that this applies to subprocessors as well. Tenancy is multi tenant with logical workspace separation and enforced role based access. Not located as of 29 Aug 2026: where data is stored as distinct from where it is processed, and what changes between tiers.

Certification is real and stated, short of accessible evidence. SOC 2 Type II and ISO 27001 are stated as maintained, ISO 42001 is referenced for the AI management system, FedRAMP In Process status is published for Westlaw and Practical Law with authorization stated as an investment goal for this product, and a corporate trust center publishes the information security management framework. What is missing is what a reader can actually pull: no attestation report, no coverage period, no audit scope, and no named auditor was located as of 29 Aug 2026, and no route to request the reports without a sales conversation was located either. Certifications are also stated at the parent company level rather than scoped to this product, which is the distinction a buyer's security team asks about.

SOC 2 Type II attestation and ISO 27001 certification with the auditor named as Schellman, renewed annually, and the 2026 cycle announced publicly. Penetration testing and red teaming partners are named as NCC Group and Bishop Fox. Certified under the EU US Data Privacy Framework. A live trust portal at trust.harvey.ai carries the current reports. Reports sit behind a portal request rather than an open download, which is a request flow rather than a sales call.

The supply chain is partly disclosed and the disclosure is unusually specific. The vendor names Anthropic's Claude Agent SDK as the architecture of the current release, names OpenAI and Google as third party AI partners in a published FAQ, and states the contractual position imposed on them. Not located as of 29 Aug 2026: which model serves which task, where the models run, a maintained subprocessor list for this product, and any commitment to notify customers before the model supply chain changes. That last gap is material here because the product was re architected onto a different vendor's agent framework within the past year.

Publishes a subprocessor list naming model and infrastructure providers including OpenAI, Anthropic, Google Cloud, AWS and Microsoft, alongside a subprocessor update FAQ and a security diagram showing model access through Bedrock and Vertex AI. States zero data retention and ephemeral processing at the model providers. Not located as of 29 Aug 2026: a published commitment to notify customers before the model supply chain changes, as distinct from an FAQ describing a change that has already happened.

The shape is visible and the number is not. Tier names and feature splits are published and reachable from the main navigation: CoCounsel Essentials, Westlaw Advantage with CoCounsel Essentials, Practical Law Dynamic Tool Set with CoCounsel Essentials, and CoCounsel Legal, each with its own plans and pricing page and segment specific variants for law firms, corporations and government. Third party analyses report that the pages run a configurator returning a figure for firms up to ten attorneys and route larger firms to a sales quote. The configurator could not be verified directly on 29 Aug 2026 because that subdomain blocks automated access, so no vendor published figure is recorded here and this grade rests on the tier structure being public while the number was not confirmed. The unit of charge is per user per month by third party report rather than vendor statement. Rebuttable in either direction with one screenshot of the configurator.

Checked the vendor home page, the platform and product pages, the help center and the trust center on 29 Aug 2026. No pricing page, no published rate, no stated unit of charge and no published tier structure located. Access to pricing runs through a demo request, which is sales gated and earns no credit. Third party per seat estimates exist in trade coverage but are not vendor published and do not move this axis.

Who the product serves is documented precisely, across more segments than any other record on this index. Published navigation splits law firms under ten attorneys from firms over ten attorneys, and separately addresses corporations, government including courts, prosecutors, public defenders, city attorneys and law enforcement, and academic use across more than 200 law schools. Practice area coverage is enumerated with its own pages spanning business and commercial, civil litigation, criminal defense, family, labor and employment, mergers and acquisitions and personal injury. The limits are stated structurally rather than in prose: the tier table shows exactly which content each plan reaches, so a buyer can see where coverage stops.

Describes its segments with substance: large law firms first, expanding into corporate legal departments and professional services, with 700 plus customers across 58 plus countries and practice coverage spanning litigation, transactional diligence, regulatory and tax. Not located as of 29 Aug 2026: a statement of the boundaries, meaning which firm sizes or practice areas the product is not built for, which is what the A bar asks for.

A published FAQ answers the question directly and in the negative, stating that user content and prompts are not used to train or improve the product or the underlying models, and that third party AI partners including OpenAI and Google are contractually prohibited from using customer data to train their models. The vendor also states it has established controls to turn off those third parties' abuse monitoring so there is no human access to customer data. The commitment as located sits on a public FAQ and trust page. No matching term was located in a published customer agreement as of 29 Aug 2026, and the subscription agreement is not published.

The vendor security page states that by default it never trains on customer data and that it contractually prohibits model providers from training on customer data. A subprocessor FAQ states customer data is never used to train models unless explicitly authorized by both the customer and the vendor. The same page defines customer data as uploaded documents and customer content as queries and responses as separate contractual terms. No matching term was located in a published agreement as of 29 Aug 2026.

REGRADED 29 Aug 2026 after the value set was amended; previously recorded at disclosed without a period, following an earlier correction on the same row. The vendor publishes a direct answer in a product FAQ: retention of user content and prompts is customised to the customer, set at the organisation account level with the support of a client manager, and the vendor deletes in accordance with the customer's own retention policy. Users can additionally delete their content and prompts in the product at any time. That is real customer control over the window and it is now recorded as such. Two qualifications kept it below the top value and are worth a buyer's attention: no retention period is published, so the window is whatever the customer negotiates rather than something disclosed in advance, and setting it runs through a client manager rather than an administrator control the customer operates alone. The separately published zero retention statement covers the model provider API layer rather than the product's own storage.

REGRADED 29 Aug 2026 after the value set was amended; previously recorded at disclosed without a period, which understated real customer control. The security page states that customers determine what data to upload, how long it is retained, and whether it can be shared internally, and help centre documentation covers configuring vault retention settings including triggers and deletion timelines. That is retention configured by the customer inside the product, which is the strongest form of the control this value describes. Recorded at customer controlled rather than the top value because no zero retention setting for the vendor's own storage was confirmed in public material as of 29 Aug 2026, and no default period is published, so a customer knows they can set the window without knowing what it is before they do. Zero data retention is stated separately as a requirement imposed on model providers, which is a different layer.

Searched the product page, the AI hub, the corporate trust center, the legal blog and the published integration announcements on 29 Aug 2026. Integrations with iManage, NetDocuments and SharePoint are named, and corporate identity management integration is stated, but no vendor material was located addressing whether retrieval enforces source system permissions at query time per user, or how ethical walls and matter level segregation are handled. A third party review refers to granular roles and ethical walls, which is not vendor material and was not treated as evidence.

Help center documentation states the product follows existing NetDocuments permissions, that a user sees only the cabinets, matters, folders and files they can already access, and that the product does not expand or modify permissions. The iManage integration is documented as a direct OAuth connection that respects iManage permissions and ethical walls. Separate admin documentation covers connecting, syncing and monitoring the firm's own ethical walls provider.

Searched the product page, the AI hub, the corporate trust center and the corporate terms of use on 29 Aug 2026. No clause addressing government or law enforcement requests for customer data was located for this product, and no transparency report was located. Noted for a future reader: a commitment to inform the customer of third party data access requests, audited under ISO 27701, was located on the trust center for a different Thomson Reuters product line, ONESOURCE Pagero. It was not treated as covering this product and does not support a value here.

Checked the security page, the published security addendum and the subprocessor update FAQ on 29 Aug 2026. No located term or policy addresses government or law enforcement requests for customer data, and no transparency report was located.

The corpus is named consistently and prominently: output is grounded in Westlaw and Practical Law, both proprietary databases the vendor owns and sells separately, maintained by attorney editors, with the editorial population stated at more than 1,200 full time attorney editors. Vendor material is explicit that the system draws on this content rather than scraping the public internet. The rights basis is ownership rather than a stated licence, and no update cadence or lag for the underlying corpus was located as of 29 Aug 2026.

The published subprocessor material identifies RELX and LexisNexis as a source provider behind an Ask LexisNexis feature, alongside web search providers, and product material refers to premium legal databases and curated public sources. The identification appears in the subprocessor list rather than a coverage page. No licence or rights basis, jurisdiction list or update cadence for the primary law corpus was located as of 29 Aug 2026.

Treatment signals come from KeyCite, the vendor's own named commercial citation service, and appear with the authority. Vendor engineering material states that reports carry inline citations, KeyCite flags and excerpts, and that the research agent uses KeyCite warnings to flag questioned or overruled cases and adjusts its strategy accordingly. Separate vendor material describes the system recognising red and yellow validity flags and factoring them into its analysis rather than only displaying them. First value on this signal in the index that rests on a licensed commercial citator.

Checked product pages for the research module, the help center and the subprocessor material on 29 Aug 2026. A LexisNexis sourced research feature is documented, but no public material was located addressing whether authority returned by the product carries a treatment signal or is checked for subsequent history.

Searched the product page, the AI hub, the legal blog, the engineering blog and the trust center on 29 Aug 2026. Vendor material describes transparent research logs, step by step reasoning, an audit trail and prompting guardrails that restrict output to a defined set of validated content. None of that states what the product does when it cannot ground an answer. No explicit no answer or abstention path was located, and no confidence or grounding score exposed to the user was located either.

The vendor publishes measured hallucination rates and describes how hallucinated claims are detected and scored. Checked that research material, the product pages and the help center on 29 Aug 2026 and did not locate published material describing an explicit no answer or abstention path when the product cannot ground an answer.

A court record exists in which a predecessor brand of this product is mentioned. In Fletcher v. Experian Information Solutions, No. 25-20086, the Fifth Circuit sanctioned appellate counsel $2,500 on 18 Feb 2026 after a reply brief was found to contain 16 fabricated quotations and 5 further misrepresentations of law or fact. The court's findings address counsel's conduct: failure to verify, and evasive and misleading responses to the show cause order, which the court treated as conduct unbecoming under FRAP 46(c). The products counsel identified as the possible source were named by the court as Casetext and vLex, described as generative AI products. Casetext is the brand Thomson Reuters acquired in 2023 and no longer sells under its own name; this record is filed here because CoCounsel is the surviving product. The court made no finding about either product's own output. Instrument also searched: the AI Hallucination Cases database maintained by Damien Charlotin.

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 sanctions summaries from Norton Rose Fulbright covering 2026 and two vendor maintained 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.

Searched the product page, the AI hub, the legal insights hub, the legal blog and the published Fiduciary-Grade AI material on 29 Aug 2026. The vendor publishes a substantial named framework and a buying guide addressed to professional standards in general terms, and markets a responsible AI position to courts and government. No engagement with any named ethics opinion was located, including ABA Formal Opinion 512 and state bar guidance.

Published material refers in general terms to aligning with the high standards expected of legal work and to designing the product so that verification is easy. Checked the blog, resource pages and help center on 29 Aug 2026 and did not locate engagement with any named ethics opinion, including ABA Formal Opinion 512 or state bar guidance.

Vendor material leads with time compression and quantifies it: a 61 percent reduction in time to draft correspondence, a customer story reporting a task that took an hour completed in five minutes or less, and a commissioned return on investment study. Searched the product page, the AI hub, the legal blog and the customer stories 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. The savings figures are published without anything on the client's side of the equation.

Vendor material offers impact and return on investment resources framed around what the product does for a firm or business, and the help center documents usage analytics dashboards and reporting APIs. Checked those surfaces on 29 Aug 2026 and did not locate a per matter record of AI assisted work intended for fee purposes, or any published guidance on billing, fee or client disclosure treatment.

A published FAQ names the third party model providers that may process customer content, OpenAI and Google, and states the contractual prohibition imposed on them, which is the substance of a model provider disclosure and is available with no agreement in place. A corporate trust center publishes the information security framework. What was not located as of 29 Aug 2026 is a maintained subprocessor list for this product, and any client facing consent or notification pack a firm could forward to its own client. Recorded at the model provider disclosure level rather than the full pack.

UPDATED 29 Aug 2026 during the trust portal sweep; value unchanged, evidence enumerated. The trust centre was opened directly and its published inventory is the most complete outside counsel readiness pack on this index. Available without a request, as named items: a Data Processing Addendum, a Business Associate Addendum, a Data Subject Requests item, completed self assessment questionnaires in three standard formats being CAIQ v4.0.3, SIG Core and SIG Lite, a Data Flow Diagram, a Network Diagram, a HIPAA report, a report titled Security and Privacy of Customer Data, and a Security Welcome Packet. A Subprocessors section is published as a standing part of the trust centre. Sensitive documents sit behind a self serve access request with a bulk download option. Compliance items are listed individually and include statements of applicability for ISO 27001, 27701 and 42001, which tell a client's reviewer what each certification actually covers. A firm answering a client AI clause could assemble a complete response from this without a sales conversation. One limitation recorded honestly: the subprocessors list renders client side and its contents were not retrieved in this pass, so the section's existence is established rather than the identity of the subprocessors in it.

Several elements of a disclosure record are available and documented. Vendor material states that each research report includes transparent research logs with verifiable citations and step by step reasoning, that sources are hyperlinked with highlighted excerpts in a dedicated tab, and that KeyCite validity flags appear with the authority, which together evidence what was retrieved and its status. Two elements are missing: the model used is not identified per output, since the vendor states the system is model agnostic across multiple providers, and no per document export covering model, sources and human verification together was located as of 29 Aug 2026.

Published material documents audit logs as a default enterprise control, inline links from assertions to the specific source passages behind them, and usage analytics available through a dashboard and APIs. Checked those surfaces on 29 Aug 2026 and did not locate a per document export covering model used, sources retrieved and human verification together.