Relativity vs Reveal: how they compare in 2026

The models are the engine of a core capability layered on a platform that would function without them. RelativityOne is an end to end ediscovery system covering legal hold, preservation, collection, processing, review, production and analytics, and it has sold for two decades. The aiR suite of five generative products sits on top of that and drives substantial capability rather than peripheral features, which is why this is not a C. Worth recording precisely because it distinguishes this record from the AI native vendors: technology assisted review has existed in this platform for years, so the generative layer is the newest of several model based capabilities rather than the first, and third party analysis characterises the aiR features as evolutionary within that lineage. Sixth B on this axis.

The first A on this axis for an ediscovery platform, and it is earned on company structure rather than marketing. Reveal was assembled by acquiring model capability and building the platform around it: NexLP and its data science team became Reveal AI, Brainspace brought visual analytics and concept search, and Reveal 11 was released explicitly as the integration of Brainspace, Reveal AI and Reveal Review into one platform. The model layer is not a feature on top of a review tool, it is the thing the review tool was assembled around. Capability spans supervised and unsupervised learning, deep learning and natural language processing, with an AI Model Library of more than 30 reusable pre built models, customer trained custom models, predictive scoring, concept clustering, translation and transcription, sentiment analysis, and generative review on aji, the GenAI review engine generally available since 30 September 2025, alongside the earlier Ask feature. Corrected 29 Aug 2026: the original note named Ask as the generative capability and did not mention aji. The grade is unaffected and if anything better supported, since aji is a purpose built engine rather than a bolt on. Distinguished deliberately from Everlaw and Relativity, both graded B, where a document review platform predates and stands without the model layer. Here removing the models removes most of the product.

Grounding is real and documented through an explainability mechanism, short of published measurement. The vendor states that aiR for Review surfaces impactful content backed by transparent rationale and that aiR for Privilege explains every decision, so each output carries a stated basis a reviewer can inspect against the document rather than a bare classification. The suite is described as designed to be transparent, reviewable and defensible, with safeguards derived from published AI Principles. That is a documented verification surface tied to specific documents. Searched the aiR product pages, the artificial intelligence overview, the corporate data solutions pages and the learning centre on 29 Aug 2026 and located no accuracy figure, no precision or recall number, no hallucination rate, no test set, no published evaluation methodology and no independent benchmark participation. A partner published case study describes predictions as highly accurate, which is a customer's characterisation rather than a measurement and was not treated as evidence.

Grounding is real, documented at product documentation level, and paired with a customer side measurement loop, short of any vendor published accuracy figure. Corrected 29 Aug 2026: the original note assessed this axis on Ask alone and never reached aji, Reveal's GenAI review engine, which had been generally available since 30 September 2025. aji produces in document reasoning for every generative rating and in text citations on positively rated documents, which the vendor documentation frames as opening the AI black box. It also ships a two stage measurement workflow: a Calibration Review that compares the reviewing attorney's own coding decisions against generative ratings on the same sample and reports an Agreement Rate, with a low rate returning the user to refine the definition, followed by an optional Validate stage run against a larger random set drawn outside the calibration sample to confirm ratings hold. A named agreement metric against attorney judgement on a random holdout is a real measurement instrument placed in the customer's hands and is better than anything else located in this category. Held at B rather than raised to A because every figure produced is produced by the customer on their own matter and none is published: searched the platform pages, the why Reveal page, the Ask resource page, the aji product pages, the aji documentation at docs.revealdata.com, the blog and the trust centre on 29 Aug 2026 and located no vendor published accuracy figure, no precision or recall number for the pre built models, for aji or for Ask, no hallucination rate, no published test set and no independent benchmark participation. The vendor describes its predictive scores as accurate without publishing a number.

A real published commitment with documented review surfaces and an explicit control philosophy, short of published thresholds. The vendor's AI Principles state the aim of technology that is clear, fair and gives customers the utmost control, and the aiR suite is described as transparent, reviewable and defensible. The review surface is concrete: every decision carries a rationale, and aiR for Privilege predictions are positioned to guide counsel's second pass review rather than to replace it. The vendor also invests in operator competence in a way no other record here does, running a certification programme covering generative AI and individual aiR products, with published guidance on building, testing and trusting prompts. Third party analysis states the product does not make autonomous privilege designations and that a human reviewer still makes every privilege call, which corroborates the position without being vendor material. Not located as of 29 Aug 2026: a published threshold at which a document routes to a human, and what the vendor commits to when an output is wrong.

A real published commitment with genuine control surfaces, short of thresholds. The control model is unusually concrete because so much of the AI is customer trained: a user builds custom models as easily as creating a tag, trains them through AI driven batches that select documents intelligently, evaluates effectiveness after each training round, and uses interactive simulations to decide when training is sufficient. That places the decision about whether a model is good enough with the customer and gives them the instrument to make it. Predictive scores prioritise review rather than replacing it, and Ask returns supporting materials alongside its narrative answer. Not located as of 29 Aug 2026: any published threshold at which a model is considered fit to rely on, what Ask does unaided, and what the vendor commits to when an output is wrong.

Named customers and a specific, checkable adoption figure, short of vendor published outcome numbers. Adoption is quantified precisely rather than vaguely: 192 of the Am Law 200 firms and more than 110 legal service provider partners use RelativityOne, which is a figure a reader can test against the published Am Law list. Named organisations are published as aiR success stories including Alvarez and Marsal, Cimplifi and Gilbert and Tobin, with Alvarez and Marsal stated to have used aiR for Review, aiR for Privilege and aiR for Case Strategy on a single complex matter. Quantified results exist but sit in partner published material rather than vendor material: a Relativity Gold Partner published a case study covering 30,000 documents analysed with aiR for Privilege and 610 hours saved in privilege review, with auto generated log descriptions replacing a manual process. That is a named scope, a named figure and a described method, and it is recorded here as partner published rather than treated as the vendor's own evidence.

Partner and testimonial material stands where deployment evidence would go. A service provider partner publishes an account of the full platform in use covering legal hold, processing, early case assessment, review, Reveal AI and Brainspace, and an attributed practitioner testimonial is published on the vendor site. Corporate history is documented with specifics that are checkable, including a $250m investment from K1 Investment Management and a sequence of named acquisitions. But none of that is deployment outcome. Searched the platform pages, the why Reveal page, the blog and the resources index on 29 Aug 2026 and located no named customer paired with figures and a date, no case study with an assessable method, and no adoption count comparable to those published by the other two ediscovery vendors on this index.

Substantive published commitments including one control no other record on this index discloses, short of the training and retention limbs. The distinctive element is specific and consequential: the vendor states it has opted out of the abuse and harmful content monitoring offered by Microsoft, so that no unauthorised users have access to raw inputs or outputs. Abuse monitoring is the standard route by which provider staff may review customer prompts, and opting out of it is the single most concrete confidentiality decision disclosed anywhere in this pull. Alongside it: a privileged access management solution and a classification schema dictating how confidential data is handled, and a vendor risk management team reviewing Azure's security and privacy posture at least annually to validate controls are operating effectively. Two gaps hold this off an A. No statement on whether customer content may be used to train models was located, either at the vendor or provider layer. No retention or deletion terms were located. Both matter for a platform holding entire document universes for live matters.

Substantive published commitments with a dedicated document on the generative feature, short of the training limb. The vendor publishes a security and data privacy white paper specifically addressing the Ask feature and the use of generative AI, which is a targeted disclosure few vendors here produce, and states a commitment to the confidentiality and integrity of customer data backed by ISO 27001 and SOC 2 Type 2 certification. Private in firewall deployment is offered as an architectural answer to confidentiality, letting an organisation keep discovery data inside its own perimeter with deep integration to internal authentication and logging. Granular audit logging and role based access controls are published. What holds this off an A is that the training question is not answered on the surfaces reached: the vendor's own guidance tells buyers to ask whether their documents are used to train third party AI models, and no statement of Reveal's own position on that was located outside the gated white paper, which requires a form submission. Retention terms were not located either.

The audience is professional and the position is unstated. Users are law firms, corporate legal departments, government and regulatory response teams, and legal service providers, with no consumer surface located, and the product is a review and investigation platform rather than an advice tool, so the advice line question arises less sharply than for a research or drafting product. Searched the aiR product pages, the artificial intelligence overview, the data solutions pages and the learning centre on 29 Aug 2026 and located no published position on advice versus tooling, no treatment of competence or supervision duties as professional obligations, and no jurisdiction limits. Worth recording as adjacent rather than as credit: the vendor runs a substantial user certification programme, which addresses operator competence as a commercial and training matter rather than as the professional duty it also is.

The audience is professional and the position is unstated. Users are law firms, corporate legal and compliance teams, government and alternative legal service providers, with no consumer surface located, and the product analyses collected evidence rather than giving advice, so the advice line question arises less sharply than for a research or drafting tool. Searched the platform pages, the why Reveal page, the blog and the Academy pages on 29 Aug 2026 and located no published position on advice versus tooling, no treatment of competence or supervision duties as professional obligations, and no jurisdiction limits despite translation and transcription across multiple languages implying cross border work. Recorded as adjacent rather than credit: Reveal Academy runs user certification courses across platform functions, which addresses operator competence commercially rather than as the professional duty it also is, the same pattern recorded for Relativity.

A published governance framework with real substance, short of testing results, a named owner and any bias disclosure. Relativity AI Principles are published as a standing document and the vendor states they guide everyday decision making toward technology that is clear, fair and gives customers the utmost control, with aiR safeguards described as inspired by them. The vendor states directly that it recognises the value AI can create along with its risks and commits to processes that are thoughtful, disciplined and trusted, which is an acknowledgement of risk rather than an unqualified capability claim. Governance is also operationalised through vendor risk management, with an annual in depth review of the underlying platform provider's security and privacy posture. Not located as of 29 Aug 2026: an AI management certification such as ISO 42001, published pre release testing results, a named accountable owner for model governance, and anything on uneven output across matter types, parties or populations. The word fair appears in the principles without any published work behind it, which is the gap this axis exists to mark.

A published pledge without a governance mechanism behind it. The vendor states three pillars of an AI pledge, Trust, Knowledge and Security, and publishes buyer guidance on choosing generative AI tools that tells readers to ask whether their documents train third party models, how access controls and audit logging are implemented, how data residency is handled and what certifications a vendor holds. Publishing the questions a buyer should ask its vendors is a real editorial contribution and is unusual. But a pledge of three words is not a framework, and searched the platform pages, the why Reveal page, the blog and the trust centre on 29 Aug 2026 and located no published AI governance framework document, no AI management certification such as ISO 42001, no named owner of model governance, no pre release testing regime, and nothing on uneven output across matter types, parties or populations. The last is a live question for this vendor specifically: it ships more than 30 pre built models for categorising documents by human behaviour including harassment and discrimination, and publishes nothing about how those models were validated or how they behave across populations.

CORRECTED 29 Aug 2026 during the trust portal sweep. Previously graded C because access control appeared to be the only well covered element. That reading came from marketing surfaces without reaching the dedicated trust site, and understated the published controls. Now located and published: customer managed encryption keys, so a customer can hold its own key material; Customer Lockbox, a default on control restricting the vendor's own system administrators from accessing customer workspaces unless the customer explicitly grants it, which is a strong and specific limit on vendor side access; client domains providing data separation between clients; Security Center, a monitoring application shipped to customers; SIEM integration giving customers access to their own security logs; and round the clock monitoring by the named in house security team, Calder7. Previously recorded and still standing: privileged access management, a classification schema for confidential data, the opt out from Microsoft abuse and harmful content monitoring, and annual vendor risk review of Azure. That is a substantive published policy covering most of what this axis asks. Not located as of 29 Aug 2026: a stated retention period or deletion control for customer data, prompts or aiR outputs, and a named subprocessor list. Those two absences are what hold this at B rather than A.

Substantive published policy covering most of the ground, with a trust centre providing the route to the rest. Published: ISO 27001 certification with a 2025 audit completed and the certified products listed, SOC 2 Type 2 certification with documentation available to enterprise clients, CSA STAR and GDPR compliance, granular audit logging and role based access controls built into the platform architecture, end to end encryption, support for data residency requirements in regulated jurisdictions, and private in firewall deployment. A SafeBase trust centre is published at a stable URL carrying the security posture and a documentation request route, and its published control set includes breach notification processes and procedures defined and implemented, and policies reviewed and updated at least annually. Not located as of 29 Aug 2026: a stated retention period or deletion control for customer data, prompts or Ask outputs, and a named subprocessor list.

Searched the aiR product pages, the artificial intelligence overview, the corporate and data solutions pages and the learning centre on 29 Aug 2026. No published indemnity, liability cap, carve out, warranty on output or insurance position was located, and no customer agreement or master terms was located on the surfaces reached. Recorded as a pure absence on those surfaces. The shape is worth naming for this vendor specifically: aiR for Privilege exists to reduce the risk of inadvertent production of privileged material, which is among the most consequential errors in litigation, and nothing published addresses who bears the loss if a privileged document is produced on the strength of an AI prediction. Rebuttable with one link.

Searched the platform pages, the why Reveal page, the blog, the resources index and the trust centre on 29 Aug 2026. No published indemnity, liability cap, carve out, warranty on output or insurance position was located, and no customer agreement or master terms was located on the surfaces reached. Recorded as a pure absence on those surfaces. Worth noting the vendor does publish a commercial commitment adjacent to this, that it works to make costs clear and predictable and explicitly frames the problem of an unexpected bill reaching a partner or client, so it has thought about commercial risk to the customer without addressing liability for output. Rebuttable with one link.

Real integrations exist, are named individually, and target the systems evidence actually lives in. Named as out of the box integrations for collections: Microsoft, Google, Slack and Box, with in place preservation from what the vendor calls the top productivity platforms, which is the integration that matters most for defensible legal hold. A published .NET Platform API for RelativityOne is documented through the learning programme, and the platform is explicitly built for shared working across internal teams, outside counsel and service providers, with more than 110 legal service provider partners in the ecosystem. Not located as of 29 Aug 2026: a consolidated integrations index page, per integration documentation of what moves in which direction and what an administrator configures, and legal document management connectors such as iManage or NetDocuments.

Integration is described at architectural level without named connectors. The strongest published statement concerns private deployment, where the vendor states the architecture allows deep integration with internal authentication controls, logging frameworks and enterprise security policies, which is integration at the identity and audit layer rather than the file layer and is a real capability. Processing handles more than 900 file formats, which reduces the ingestion problem a connector would otherwise solve. Searched the platform pages, the why Reveal page, the blog and the resources index on 29 Aug 2026 and located no integrations index page, no named connector for collection sources such as Microsoft, Google, Slack or Box, and no API documentation. That contrasts directly with the other two ediscovery records on this index, both of which name collection connectors individually.

The hosting platform is named and residency is not addressed. RelativityOne is stated to be built on Microsoft Azure, and the vendor adds a substantive point about that choice, that it is the same platform chosen by global regulators, alongside an annual vendor risk review of Azure's security and privacy posture. Naming the hosting provider and evidencing ongoing oversight of it is more than several records here manage. Searched the aiR pages, the artificial intelligence overview, the data solutions pages and the learning centre on 29 Aug 2026 and located no named regions, no customer selectable residency, no tenancy model, and no statement of where processing happens as distinct from where data is stored. For a platform serving 192 of the Am Law 200 across international matters, published residency options would be expected and none was located on the surfaces reached.

The strongest deployment position in the ediscovery category and among the strongest on the index. Three models are published and the differences between them are real: cloud, hybrid, and private in firewall deployment where discovery data stays inside the organisation's own perimeter. Private deployment is not a legacy option being quietly maintained but an area of active investment, with the vendor announcing it expanded that investment by 50 percent on stated enterprise demand and publishing architectural material on how it works, covering secure infrastructure, scalable performance, consistent compliance inside the firewall and deep integration with internal authentication and logging. Data residency is addressed directly as support for residency requirements in regulated jurisdictions, and the vendor's own buyer guidance tells readers to insist on data locality options because data sometimes must stay within a region or country. Short only of naming the available cloud regions and stating the tenancy model for the hosted option, neither of which was located as of 29 Aug 2026.

CORRECTED 29 Aug 2026 during the trust portal sweep. Previously graded C on the finding that the vendor claimed several industry certifications without naming any. That was wrong. It came from reading product and solutions marketing pages without reaching the dedicated trust site at relativity.com/trust and its compliance and privacy page, where the certifications are named individually and with versions. Published there: ISO/IEC 27001:2022 certification, ISO/IEC 27018:2019 certification, FedRAMP Moderate ATO, HIPAA compliance, IRAP assessed at Protected, a SOC 2 Type II report, a SOC 3 report, and a Cloud Security Alliance CAIQ, alongside a published request route for compliance certificates. Naming ISO 27001 at the 2022 revision and 27018 at 2019 is precise, and the set is unusually broad, spanning the US federal authorisation, the Australian government assessment at Protected level, a healthcare framework and a standardised cloud control questionnaire. A SOC 3 report is a public summary report, which is a materially more open disclosure than SOC 2 alone. The vendor also names its in house security team, Calder7, publishes a Security Center monitoring application to customers, and offers SIEM integration giving a customer access to their own security logs. Short only of a published coverage period, report date and named auditing firm, none of which was located as of 29 Aug 2026.

Certification is named, current and reachable through a published portal, short of scope and auditor detail. Named: ISO 27001, SOC 2 Type 2, CSA STAR and GDPR compliance. Currency is evidenced rather than assumed, the trust centre announcing completion of the ISO 27001 audit for 2025 and listing the products certified, which addresses the question of what the certificate actually covers more directly than most vendors do. A SafeBase trust centre is published at a stable URL as a self serve route to request security documentation, and SOC 2 Type 2 documentation is stated as available to enterprise clients. Under the three tier test that is a request flow rather than a sales gate. Not located as of 29 Aug 2026: the SOC 2 audit coverage period, the named auditing firm for either certification, and the ISO 27001 certificate expiry. Compare Relativity, held at C in the same category for claiming certifications without naming any.

The provider is identified, where the models run is stated, and the terms binding the relationship are described, short of naming the models themselves. Microsoft Azure is named as the platform the product is built on, and the vendor's disclosure about opting out of Microsoft's abuse and harmful content monitoring identifies Microsoft as the party that would otherwise have had access to raw inputs and outputs, which locates the generative processing in the Microsoft stack more precisely than most vendors manage. The relationship is governed rather than assumed: a vendor risk management team performs an in depth review of Azure's security and privacy posture at least annually to validate controls are operating effectively. Not located as of 29 Aug 2026: which specific models serve which aiR product, any subprocessor list beyond the platform provider, and any commitment to notify customers when the model supply chain changes.

The architecture is described in detail and the third party components are not identified. What is published is substantial on the vendor's own layer: models built in house through the Reveal AI lineage from NexLP, a proprietary model library, customer trained custom models, and Brainspace analytics, all of which a buyer can understand as owned rather than rented. What is not published is what sits behind the Ask generative feature. Searched the platform pages, the why Reveal page, the Ask resource page description, the blog and the trust centre on 29 Aug 2026 and located no named large language model or provider, no statement of where the generative processing runs, no subprocessor list, and no commitment to notify customers when the supply chain changes. The vendor's own buyer guidance instructs readers to ask whether their documents are used to train third party AI models, which implies third party models are in the picture without saying whose.

A real commercial term is published without a rate, and the term itself is unusual enough to record. The vendor states that aiR for Review and aiR for Privilege are included in the standard pricing and packaging for RelativityOne, so a buyer learns that two of the five generative products carry no separate charge, which is a meaningful commercial fact and one almost no vendor on this index discloses about its AI features. Flexible pricing models are referenced without being enumerated. Searched the aiR pages, the data solutions pages and the corporate pages on 29 Aug 2026 and located no rate, no unit of charge, no tier structure, and no published packaging for the remaining aiR products, aiR for Case Strategy, aiR for Data Breach Response and aiR Assist. Recorded at C on the strength of the published inclusion statement.

Checked the platform pages, the why Reveal page, the blog and the resources index 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 request. One published commercial statement was located and is recorded here because it addresses the buying experience without giving a figure: the vendor states it works with customers to make costs as clear and predictable as possible, framing the problem explicitly as the unexpected bill a practitioner has to take to a partner or client. That is an acknowledgement of the cost predictability problem in ediscovery rather than a disclosure of price, and it does not lift the grade.

Segment and practice coverage is described with substance and quantified where it can be. Segments named: law firms with a stated 192 of the Am Law 200, corporations, government, and more than 110 legal service provider partners, with the platform explicitly designed as shared working space across internal teams, outside counsel and providers. Practice coverage is enumerated by product rather than claimed broadly, five aiR products each with a distinct purpose spanning document review, privilege, case strategy, data breach response and assistance, plus legal hold, preservation, collection, processing, production and analytics. Regulatory response is named down to the agency: EPA, DOJ, FDA, SEC and third party subpoenas. Not located as of 29 Aug 2026: firm size segmentation below the Am Law tier, jurisdictional or language coverage, and any statement of what the platform is not built for.

Practice coverage is described with real substance and segment coverage more thinly. Practice scope is stated end to end across the EDRM from preservation and collection through processing, early case assessment, review, analytics and production, with more than 900 file formats supported, translation and transcription across multiple languages, and a localisable interface. Investigation types are enumerated concretely through the model library, with more than 30 pre built models aimed at recurring matters and harassment and discrimination named specifically, which is a more granular statement of what the product is for than most records here provide. Segments addressed include law firms, corporate legal and compliance, and alternative legal service providers, with the vendor noting customers may run work themselves or through a provider. Not located as of 29 Aug 2026: firm size segmentation, jurisdictional coverage stated as such, adoption figures, and any statement of what the platform is not built for.

Searched the aiR product pages, the artificial intelligence overview, the corporate and data solutions pages and the learning centre on 29 Aug 2026. No located material states whether customer content may be used to train models, either at the vendor layer or by the underlying platform provider. Recorded as silent under the rule that a value is never inferred from the absence of a contradiction, and specifically not inferred from the published opt out of Microsoft abuse and harmful content monitoring, which stops provider personnel accessing raw inputs and outputs and is a different question from whether anything trains on them. Notable as an absence given how specific this vendor is elsewhere about its data handling decisions.

Searched the platform pages, the why Reveal page, the blog, the resources index and the trust centre on 29 Aug 2026. No located material states whether customer content may be used to train models, either by the vendor or by any third party model provider. Recorded as silent under the rule that a value is never inferred from the absence of a contradiction. The absence is conspicuous rather than neutral here, because the vendor's own published buyer guidance instructs readers to ask a prospective vendor whether their documents are used to train third party AI models, listing it among the questions that separate a serious tool from a risky one, and does not answer that question about itself on the surfaces reached. A dedicated security and data privacy white paper on the Ask generative feature is published but sits behind a form submission and was not retrieved, so the answer may exist there.

Searched the aiR product pages, the artificial intelligence overview, the corporate and data solutions pages and the learning centre on 29 Aug 2026. No public material on these surfaces states how long prompts, aiR outputs or rationales are retained, whether a customer controls the window, or whether deletion is available. The gap has a particular edge for this product: aiR generates a rationale for every decision and auto generated privilege log descriptions, so the system produces a substantial body of derived commentary about a customer's documents, and nothing located governs how long that commentary persists.

Searched the platform pages, the why Reveal page, the blog, the resources index and the trust centre on 29 Aug 2026. No public material on these surfaces states how long documents, Ask prompts or generated narrative answers are retained, whether a customer controls the window, or whether deletion is available. Noted for a reader as a genuine mitigation rather than an answer: private in firewall deployment means an organisation choosing that option holds its own discovery data inside its own perimeter and therefore sets retention by controlling the store, which removes the question for those customers without addressing it for the hosted ones.

CORRECTED 29 Aug 2026 during the trust portal sweep. Previously recorded at claimed but not documented, on the basis that internal controls were asserted without customer side segregation being described. That was wrong: the product documentation describes two distinct mechanisms and describes them concretely. Client domains provide a secure way to isolate users, workspaces, groups and matters by client, with data separation such that only certified partners have access across their own clients, and client domain admins administer within that boundary. Separately, Customer Lockbox restricts the vendor's own system administrators from accessing customer workspaces unless the customer explicitly grants it, enabled by default, with system admins additionally required to belong to a group within a workspace to reach it. Isolation by matter is stated explicitly, which is the level a firm facing product needs, and vendor side access is constrained by a default on control rather than a policy promise. Recorded at own model documented rather than the positive value because the product operates its own permission structure rather than inheriting a document management system's access model at query time, and because no material was located stating that aiR retrieval and generation respect those boundaries per user when the AI runs across a workspace. Ethical walls are still not named as such. Previously recorded and still standing: privileged access management and a classification schema governing vendor handling of confidential data.

The product maintains its own documented permission model, described at a useful level of specificity. Published: granular audit logging and role based permissions built into the platform architecture rather than layered on, access controls and encryption configured within customer environments, and for private deployment, deep integration with the organisation's own authentication controls and logging frameworks, which brings the firm's identity model to bear on the platform. The vendor's buyer guidance also raises shared multi tenant infrastructure as a question a buyer should put to any vendor, and private deployment is its answer to that. Recorded at own model documented rather than the positive value because no published material states that retrieval or AI model application enforces those permissions per user at query time, and no legal document management integration was located whose access model could be inherited. Ethical walls are not named as such anywhere located.

Searched the aiR product pages, the artificial intelligence overview, the corporate and data solutions pages and the learning centre 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. Worth recording as context: the vendor markets consolidated regulatory response across EPA, DOJ, FDA, SEC and third party subpoenas, so its customers use the platform precisely to manage government demands for their own data, and what happens when a government instead demands data from the platform is unaddressed on the surfaces reached.

Searched the platform pages, the blog, the resources index and the trust centre 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. The trust centre's published control set includes defined and implemented processes for security breach notification, which addresses telling a customer about a compromise rather than about a lawful demand, and the two were not conflated.

No primary law corpus is identified because the product does not hold one. Retrieval and analysis run against the customer's own collected evidence, ingested through legal hold, preservation and collection from named enterprise sources, so the corpus is the document universe for the matter and its provenance is the discovery process. Searched the aiR pages, the artificial intelligence overview and the data solutions pages on 29 Aug 2026 and located no vendor supplied legal corpus, no licence basis and no update cadence, and none would be expected. One adjacent capability was considered and not treated as a corpus: the platform can carry coding decisions, compliance workflows and privilege calls from prior matters into new ones, which reuses the customer's own past work product rather than any vendor held material.

No primary law corpus is identified because the product does not hold one, but this vendor raises a related provenance question that none of the others do. Retrieval and analysis run against the customer's own collected evidence, so for that corpus the provenance is the discovery process itself and no vendor corpus would be expected. What is different here is the AI Model Library: more than 30 pre built models are shipped ready to identify content and categorise documents by human behaviour, and a pre trained model necessarily encodes whatever it was trained on. Searched the platform pages, the why Reveal page, the blog and the resources index on 29 Aug 2026 and located no statement of what those models were trained on, on whose data, on what rights basis, or how they are validated or updated. Recorded as not addressed on that basis rather than as inapplicable.

Searched the aiR product pages, the artificial intelligence overview, the data solutions pages and the learning centre on 29 Aug 2026. No material was located addressing whether authority carries a treatment signal or whether subsequent history is checked, and no commercial citator licence was located. Noted for context: this is an ediscovery and investigation platform whose corpus is collected evidence rather than published case law, so a citator is outside its design entirely, including for aiR for Case Strategy, which builds argument from the document record rather than from authority.

Searched the platform pages, the why Reveal page, the blog and the resources index on 29 Aug 2026. No material was located addressing whether authority carries a treatment signal or whether subsequent history is checked, and no commercial citator licence was located. Noted for context: this is an ediscovery and investigations platform whose corpus is collected evidence rather than published case law, so a citator is outside its design entirely, consistent with both other ediscovery records on this index.

Searched the aiR product pages, the artificial intelligence overview, the data solutions pages and the learning centre on 29 Aug 2026. No published material describes what the product does when the evidence does not support a determination, and no explicit no answer path was located. Two adjacent features were considered and not treated as satisfying this signal. Every aiR decision carries a rationale, which explains a determination that was made rather than declining to make one. And aiR for Privilege produces a prioritised queue surfacing high probability privileged documents, which is ranking by confidence rather than an abstention path, and no confidence threshold exposed to the user was located. For a privilege product the question of what happens on a genuinely ambiguous document is the sharpest version of this signal and it is unaddressed.

Searched the platform pages, the why Reveal page, the blog, the resources index and the trust centre on 29 Aug 2026. No published material describes what Ask does when the evidence does not support an answer, and no explicit no answer path was located. Two adjacent features were considered and not treated as satisfying this signal. Ask returns a narrative answer with supporting materials, which lets a reader check an answer that was given rather than telling them when nothing supported one. And predictive scoring exposes model confidence per document for review prioritisation, which is a ranking signal on classification rather than an abstention path on a generative answer, and the two operate on different parts of the product.

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. Note the exposure differs from a research tool: this product analyses collected evidence rather than generating citations to authority, so its characteristic failure would be a wrong privilege call or a mischaracterised document rather than an invented case, and neither would ordinarily surface in a hallucination database.

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. Note the exposure differs from a research tool: this product analyses collected evidence rather than generating citations to authority, so its characteristic failure would be a mischaracterised document or a wrong behavioural classification rather than an invented case, and neither would ordinarily surface in a hallucination database.

Searched the aiR product pages, the artificial intelligence overview, the data solutions pages, the certification programme pages and the learning centre on 29 Aug 2026. No engagement with any named ethics opinion or bar guidance was located, including ABA Formal Opinion 512 and state bar guidance. Also not located, and more surprising for this vendor: any engagement with the Federal Rules of Civil Procedure or with the substantial body of case law on technology assisted review and defensible process, which is the natural professional touchstone for an ediscovery platform marketing defensibility. The vendor publishes AI Principles governing its own conduct and a certification programme establishing operator proficiency, neither of which engages with the professional rules binding its users.

Searched the platform pages, the why Reveal page, the blog, the resources index and the Academy pages on 29 Aug 2026. No engagement with any named ethics opinion or bar guidance was located, including ABA Formal Opinion 512 and state bar guidance. Also not located, and consistent with the other two ediscovery records here: any engagement with the Federal Rules of Civil Procedure or with the case law on technology assisted review and defensible process. That is now an absence recorded across all three ediscovery vendors on this index, which makes it a category pattern rather than a vendor failing.

Savings are claimed and quantified in partner material with nothing published on the client's side of the equation. The vendor's framing is cost and time reduction: automating privilege review to increase productivity and reduce cost, identifying impactful content in substantially less time, and reducing cost and response time across matters by carrying prior coding decisions forward. A Relativity Gold Partner published a case study recording 610 hours saved in privilege review on a 30,000 document analysis. Privilege review is billed work, and 610 hours is a large number in that context. Searched the aiR pages, the data solutions pages and the learning centre on 29 Aug 2026 and located no per matter record of AI assisted work intended for fee purposes, and no guidance on billing, fee or client disclosure treatment. Recorded at savings claims only.

Savings are implied through efficiency framing and the vendor addresses cost from an unusual direction. Published claims centre on getting to relevant content faster, prioritising review through predictive scores and reducing time to insight. Separately and more distinctively, the vendor addresses cost predictability directly, stating it works with customers to make costs clear and predictable and framing the problem as the unexpected bill a practitioner has to take to a partner or client. That engages with what the client ultimately pays, which is adjacent to this signal and rarer than the usual hours saved claim, but it concerns the vendor's own invoice rather than the firm's. Searched the platform pages, the why Reveal page, the blog and the resources index on 29 Aug 2026 and located no per matter record of AI assisted work intended for fee purposes and no guidance on billing, fee or client disclosure treatment.

Some genuinely useful material is published and the artifacts this signal names are not. Available: identification of Microsoft Azure as the platform, the disclosure that the vendor opted out of Microsoft abuse and harmful content monitoring so provider personnel cannot reach raw inputs and outputs, and a statement that Azure's security and privacy posture is reviewed in depth at least annually. A firm could forward the abuse monitoring point usefully, since it answers a question client AI clauses increasingly ask. But searched the aiR pages, the artificial intelligence overview, the data solutions pages and the learning centre on 29 Aug 2026 and located no subprocessor list, no named security certification, no published data processing agreement, and no client facing consent or notification pack. Recorded as not addressed because no assembled material exists to point a client to.

A published trust centre provides the route, and the specific artifacts this signal names are not behind it on the surfaces reached. Available without a sales conversation: a SafeBase trust centre at a stable URL carrying the security posture with a documentation request flow, named current certifications including ISO 27001 with the 2025 audit completion and certified product list, SOC 2 Type 2 with documentation stated as available to enterprise clients, CSA STAR and GDPR. A firm could evidence a good deal from that. Searched the trust centre entry point, the platform pages and the blog on 29 Aug 2026 and located no subprocessor list, no statement naming which model providers see client content, and no client facing consent or notification pack. Recorded at on request on the strength of the published documentation route falling short of the named artifacts.

The most complete disclosure material of any ediscovery record here, assembled from product features rather than offered as a single artifact. Every aiR decision carries a rationale, so the basis for each determination is recorded per document rather than reconstructed afterwards. The vendor states full audit trails and documented data governance across the platform, and describes the aiR suite as designed to be transparent, reviewable and defensible, with defensibility a stated design goal rather than a marketing adjective. aiR for Privilege generates privilege log descriptions automatically, which is a court facing artifact produced as a by product of the AI work itself. Two elements are missing: no single per document export combining model used, sources retrieved and human verification was located, and no model is named in published material so the model used could not be stated. Recorded at partial record on that basis.

Several elements of a record are available, assembled from platform features rather than offered as a disclosure artifact. Granular audit logging is published as built into the platform architecture rather than added on, Ask returns supporting materials alongside each narrative answer so the basis of a generative output is traceable, and the custom model workflow records training rounds and effectiveness evaluations, which is a documented account of how a classifier reached its state. Two elements are missing: no per document export combining model used, sources retrieved and human verification was located, and no model is named in published material so the model used could not be stated for the generative feature. Recorded at partial record. Noted for a reader: predictive coding and technology assisted review have an established judicial record in ediscovery, and none of the three ediscovery vendors on this index connects its AI disclosure material to it.