Merlin Search Technologies vs OpenText eDiscovery: how they compare in 2026

The models are the engine of a core capability layered on a product that would still function without them. Alchemy's answering and analysis engines and ReviewPartner's AI reviewers sit on a full ediscovery platform with keyword and algorithmic search, review, tagging, production and multi-party workspaces that the company has sold since 2019 under the Sherlock and DiscoveryPartner names; the vendor's own framing is three engines to find and two to analyse and answer. Remove the answering models and a search-and-review platform remains. Home page, security page and launch coverage read 6 September 2026.

The models are the engine of core capabilities layered on a product that would function without them, which is the B band, and this record makes the distinction unusually easy to see because the vendor draws it itself. Its own FAQ states that OpenText eDiscovery Aviator is a suite of LLM-powered features built into OpenText eDiscovery, and lists them: key document summary, summarisation, concept label summaries, Review, Rapid Exploration and Agents. Aviator is also a company-wide brand applied across unrelated products, so it denotes an AI layer rather than a product. What that layer does here is substantial. Aviator Review reduces or eliminates first-pass responsiveness review using large language models, with sampling and token estimates before a full run, and Aviator Agents take a goal described in natural language and plan and execute the multi-step search, review and summarisation behind it. Underneath it, the platform is a complete discovery system that predates the models by many years: collection through more than 45 connectors, processing with deduplication and de-NISTing, predictive search, an integrated viewer with redaction, visualisation, audio and video review, and technology-assisted review with continuous active learning that the company states it has led on for more than fifteen years. Remove the generative layer and a buyer still has an end-to-end eDiscovery platform with a mature machine-learning review engine, which is what settles this at B rather than A. Verified 13 September 2026.

Grounding is real and documented with sourced outputs, short of a testable accuracy figure on the surfaces read. The home page states that questions return comprehensive, sourced answers and reports across the document set rather than search hits, the launch coverage describes a dual-window architecture separating analysis prompts from formatting prompts to keep formatting from interfering with document analysis, and the AI services page states that every service undergoes rigorous validation for accuracy, completeness and consistency at scale. No accuracy figure, test set or evaluation is published in the material read; the founder's published book on generative AI for discovery was not opened. The primary-authority limbs do not apply to a tool that reads the customer's own documents. Home page, AI services page and launch coverage read 6 September 2026.

Grounding is real and documented and a measured figure is published without the test set behind it, which lands in the B band. The grounding mechanism is specific and checkable by design: key document summaries are generated complete with document ID citations expressly so that users can confirm veracity, and the material the AI reads is the customer's own collected evidence rather than an external corpus, so a reviewer can always go to the underlying document. A measured figure is published and it is the right one for this product class: 90 per cent or better recall with optimised workflows, recall being the metric that matters in responsiveness review because it measures what the process failed to find. Alongside it sit 88 per cent faster review and 75 per cent cost savings against manual review, which measure effort rather than correctness. The A band asks for the test set described and the failure modes named, and neither is published. Nothing states the matter, corpus, volume or protocol behind the recall figure, and the qualifier optimised workflows does real work in the vendor's favour since it conditions the number on expert prompt engineering and proven workflows without saying what either consists of. One limitation is volunteered and is credited here because it is the closest thing to a named failure mode on the record: the FAQ states that while the model has multilingual capabilities, primary testing and validation have been on English-language documents and prompts. On a platform sold for cross-border investigations that is a material and honest disclosure. Verified 13 September 2026.

Review surfaces and constraints are published, short of the full control structure. Granular role-based access control exposes every function as an assignable permission, audit logs track each person's views, tags and searches, multi-party sites give user groups private tag sets and shared searches, and the platform delivers sourced reports a lawyer works from; ReviewPartner's AI reviewers are described as applying consistent judgment across files with validation. What is not published is the threshold at which AI review decisions stand without human sampling, what executes without approval, or a stated route back after a wrong classification. Security page, home page and AI services page read 6 September 2026.

A written commitment that the models work alongside human judgement, backed by real review surfaces and by contract, short of the full control structure, which is the B band and a strong instance of it. The commitment is contractual rather than promotional, which is rare on this axis. Clause 11.3 of the published Cloud Services Use and Delivery Terms states that use of AI Components does not replace decision-making and judgement by natural individuals, that they are intended to provide additional knowledge to support such decision-making, and that the customer remains solely responsible for decisions taken as a result of the outputs; the same clause prohibits unacceptable risk use as defined in the EU AI Act. Clause 11.1 acknowledges that customer-chosen parameters and inputs may carry assumptions, biases and limitations affecting output quality. The product surfaces match that framing: the vendor describes its own approach as GenAI human-assisted review, summaries carry document ID citations for verification, and Aviator Review is explicitly iterative, letting a team test criteria against sample data with token estimates and refine before committing to the full set, which is a genuine stopping point rather than an assurance. What the A band requires is absent. No threshold is published at which any feature acts alone or escalates, nothing states what an Agent will not attempt, and nothing describes what happens after an output is found to be wrong. R124(2) was applied to clause 11.3 and it does not qualify: it is a general statement across all AI Components rather than a constraint attached to a named mode or tier stating what that tier's output may not be used for. Verified 13 September 2026.

A named customer without figures, and figures without a named customer. Dara Tarkowski, managing partner of Actuate Law, is quoted by name on the 2023 cloud utility pricing launch about adopting the pricing model, with no measured outcome; the home page states most clients save fifty to sixty per cent on hosting costs, unattributed, and the TransPerfect partnership announcement states Reef ECA reduces document sets by over ninety per cent on average, which is the partner's figure. US Legal Support and TransPerfect are named partners rather than customers. Nothing joins a named customer to a figure. Home page, 2023 launch release and April 2026 partnership release read 6 September 2026.

Real deployment evidence with substance, short of joining the names to the figures, which is the B band in its own words. The naming is strong and spans the buyer types this product serves. The UK Serious Fraud Office is named with an attributed individual, Richard Day, quoted saying the platform transformed how the agency investigates and is embedded in its workflows, which is an unusually direct reference from a prosecuting authority. Bosch is named on the corporate side, and three law firms are named: Gleiss Lutz, Eversheds Sutherland and Pillsbury. Figures are published and they are specific: 75 per cent cost savings against manual review, 88 per cent faster document review, 90 per cent or better recall, up to 88 per cent cost savings on the product page, and a review of 48,756 documents completed in under two weeks. The problem is that the two sets do not meet. The customers who are named carry no figures, and the deployments carrying figures are anonymous, being described as a global technology leader and a major housing provider. That is precisely the split the B band names as the common failure. Nothing is dated, so a reader cannot tell when any deployment ran, and no method accompanies any figure, so the 75 per cent and 88 per cent claims cannot be assessed against what they were measured from. The individual customer story pages were not opened; under R25 they corroborate a grade that stands on the product and Aviator pages, and they are what would move this row to A. Verified 13 September 2026.

Substantive published commitments on segregation and work product, short of a training statement and an agreement. Segregation is the strongest in the pull: each client in its own AWS account with no shared database or infrastructure, each matter on its own servers, database and storage with its own URL and login, and multi-party sites with private fields, tag sets and searches that the security page describes as protecting group work products, which is work-product treatment in writing. Confidentiality controls include document lockdown with time-limited authorisation tokens, encryption in transit and at rest, and audit logging. Not located: any statement on training use, since no customer agreement is published and the privacy policy was not opened; any retention or deletion commitment; and the identity of the models the platform runs. Security page, Alchemy page and security FAQ read 6 September 2026.

Substantive published commitments across most of the ground, with two real gaps and one countervailing disclaimer, which is the B band. What is committed is unusually concrete for this axis. Training is excluded in terms: OpenText states it does not train models and that customer data, input prompts and outputs are never used to train the model. Segregation is documented at the level that matters in discovery: each LLM interaction is a separate session scoped to a single project and results are not shared across projects, and the eDiscovery terms require OpenText's conflicts check to be successfully completed before the service is provided for a matter, which is a conflicts mechanism rather than an access control. Third-party model handling is answered rather than assumed, the provider being named and session data stated to be deleted from Amazon Bedrock when the session closes, with Bedrock not storing or logging customer data. Clause 5.2 of the agreement confirms the customer owns its Content and OpenText claims no ownership, and clause 9 binds both parties to strict confidence for the term and three years after. Three things hold it at B. Privilege and work product are not addressed by name anywhere located, on the one product class where privilege review is a defined workflow. Retention and deletion of Content are deferred to Supplemental Terms under clause 6.5, which also states OpenText has no obligation to retain or delete Content except as agreed. And clause 11.4 runs the other way and is recorded because a buyer must weigh it: for AI Components using large language models, OpenText does not guarantee the protection of privacy, rights to use, or the accuracy of outputs. Verified 13 September 2026.

No advice line or supervision statement was located. The product is sold to law firms, corporate legal departments and investigators and its AI reviewers are described as applying senior attorney-level judgment, which is a capability claim rather than a position on where the lawyer's responsibility sits; no surface read states that outputs are not legal advice or how the product supports a supervising lawyer's duties, and no customer agreement exists on the site to carry such a statement. Home page, security page and AI services page read 6 September 2026.

A real published position on advice versus tooling, short of the supervision and competence dimension, which is the B band and its named common shape. The position is contractual, which sets this record apart from the two contract-lane records graded D in the same pull. Clause 11.3 of the published agreement states that use of AI Components does not replace decision-making and judgement by natural individuals, that they are intended to provide additional knowledge to support such decision-making and judgement, and that the customer remains solely responsible for any decisions taken and judgements made as a result of the outputs. The same clause prohibits unacceptable risk use as defined in the EU AI Act or per industry standards, and clause 11.6 places responsibility on the customer for identifying and complying with laws applicable to the use of AI Components in its own processes. Clause 3.2 separately makes the customer responsible for compliance with laws on the use of artificial intelligence systems. The scope of the licence reinforces it: the eDiscovery subscription is granted only in connection with the customer's legal or regulatory matters, and access may not be given to a managed service, eDiscovery or document review company or consultant without prior written consent. Marketing is consistent rather than in tension, the vendor describing its own approach as human-assisted review and publishing that agentic AI strengthens rather than replaces legal judgement. What the A band asks for is missing: nothing addresses supervision or competence, no rule of professional conduct or bar authority is named, and no jurisdiction limit is stated. Verified 13 September 2026.

Security governance is published and AI governance is not. The security page names a chief information security officer reporting to the chief executive and a Technology Governance Committee co-chaired by the CISO and COO providing executive oversight, and the AI services page states rigorous validation of every service; security governance is a different subject from AI governance under the standing rule, and no responsible AI framework, ISO 42001 or equivalent, testing results or statement about uneven output across document types or languages is published. Security page and AI services page read 6 September 2026.

The published position on AI governance is contractual rather than a governance framework, with no mechanism, owner or testing regime behind it, which lands in the C band. What exists sits in the agreement and is a risk allocation rather than a governance programme. Clause 11.1 acknowledges that a customer's chosen parameters and inputs may include assumptions, biases and limitations affecting the effectiveness, quality, relevance and accuracy of outputs, which names bias as a real phenomenon and then assigns responsibility for it to the customer, input quality being the customer's sole responsibility under clause 11.2. Clause 11.3 references the EU AI Act by prohibiting unacceptable risk use, and clause 11.6 places compliance with AI-specific law on the customer. Clause 11.4 states plainly that OpenText does not guarantee the accuracy of outputs from large language model components. So the vendor is candid about limitations and silent about what it does to manage them. What the A and B bands require is absent. No responsible AI or AI principles page for this product was located, no accountable owner for model behaviour is named, no pre-release testing regime is described, and nothing is disclosed about what any evaluation has found. Bias is named as an input problem and never addressed as an output property, which matters here because responsiveness classification across custodians, languages and document types is exactly where uneven behaviour would show. Recorded under R25: the company-level Trusted Data and AI page was not opened and is what would move this row. Verified 13 September 2026.

Substantive published policy covering most of the ground, with retention and notification not located. Access control: granular role-based access control, document lockdown by time-limited token, audit logs of every view, tag and search, private subnets with no public IP addresses, TLS and AES-256 encryption. Hosting and sub-processors: Amazon Web Services named as the infrastructure provider, with per-client accounts and regions chosen by the client; AI model providers are not named. Incident practice: the SOC 2 Type II scope covers monitoring and incident response, and the single-tenant design is described as limiting the blast radius of any incident to one environment; no customer notification commitment was located. Not located: a retention period or deletion commitment, since no customer agreement is published and the privacy policy was not opened. Security page and security FAQ read 6 September 2026.

Substantive published policy covering most of the ground, short of the full set, which is the B band. The strongest limb is the one most records leave blank, being what happens to material after processing. The Aviator FAQ describes the data flow step by step: a session is created with the model through the Amazon Bedrock API directly from the customer's own OpenText eDiscovery AWS environment, the request carries the prompt and document text, the response returns to that environment, and the session is closed with all data deleted from Bedrock, which does not store or log customer data in its service logs. Interaction is scoped per project with a separate session per feature interaction, and the model has no internet access, Bedrock hosting a deep copy. Training is excluded outright. Access to independent assurance is provided for under clause 7.4, which commits OpenText to supply summaries of third-party audit reports and certifications on written request under a confidentiality agreement. Three gaps hold it at B. No subprocessor list is published: the Data Processing Addendum takes a general authorisation to engage subprocessors, naming none, with a thirty-day objection and termination route, and AWS is named for eDiscovery only in the agreement itself. No incident or breach notification practice was located on the surfaces read. And retention of Content within the platform is deferred to Supplemental Terms under clause 6.5, with clause 10 adding that deletion following termination may be delayed up to 180 days by the third-party cloud infrastructure provider. Verified 13 September 2026.

No liability position is published. The site's page inventory, taken from the navigation and footer on 6 September 2026, carries a privacy policy, a GDPR disclosure and a contact page and no terms of service or customer agreement; the security page describes controls and invites contact for security requirements without any warranty, indemnity, cap or insurance statement. This is an absence on the vendor's surfaces rather than a retrieval limit. Security page and footer read 6 September 2026.

A real published position on liability, short of the full picture, which is the B band, and it is more specific to AI than almost anything else in this corpus. The agreement is published in full and carries a dedicated Artificial Intelligence Technologies section rather than folding AI into a general disclaimer. Clause 11.3 states that OpenText shall have no liability resulting from the creation or use of outputs, or from any decisions resulting from their use, with the customer solely responsible for those decisions. Clause 11.4 is unusually blunt and is quoted in substance because a buyer should see it: for AI Components using large language models, the nature of the technology may limit the protection of privacy, rights to use, and the accuracy of outputs, and OpenText therefore does not guarantee any of the three. Clause 11.5 flags that third-party AI components may carry additional terms accepted at the point of use. Recourse of a kind exists elsewhere in the agreement: clause 7.4 entitles the customer to summaries of SOC 1 Type II, SOC 2 Type II and ISO 27001 assurance on request under confidentiality, and clause 2.3 requires OpenText to identify the reason and expected impact of any change with a material adverse effect on functionality and to discuss mitigation. What the A band asks for is not in this instrument. No liability cap appears in these terms, which route commercial caps to the order documents and supplemental terms, no vendor indemnity of any kind was located, no service level accompanies the AI features, and no insurance position is published. Verified 13 September 2026.

One integration is named with its function and no documentation an implementer could use was read. The April 2026 partnership release states that TransPerfect's Reef ECA is deployed as the processing engine across all Alchemy cloud instances, giving a unified workflow from ingestion through review; the security page offers hosting in a client's own AWS environment. No document management, review-platform or load-file integration is described with what moves and in which direction, and no integration documentation was located. Partnership release and security page read 6 September 2026.

Real integrations exist and named connections are documented, short of the depth the A band describes, which is the B band. Collection breadth is the substantive limb and is published as a figure: more than 45 connectors feed the platform, which on a discovery product is the integration that matters most, because the systems legal work lives in are the custodian's mail, chat and file estates rather than a practice management suite. A paid add-on, OpenText Connectors for Private Cloud, is listed in the purchase table, so connector coverage is an explicit commercial dimension rather than an implicit one. The platform sits in a Legal Tech family with adjacent products a buyer can combine, being OpenText Investigation, Core Insight, Core Legal Hold and Legal Knowledge Management, and the agreement names AWS as the infrastructure for the eDiscovery Services. Audio and video are ingested and made searchable in native and transcribed form, which is an ingestion capability rather than a connector but bears on the same question. What the A band requires and was not established is depth: no individual source system is named on the surfaces read, no direction of flow is described, and nothing states what a customer must configure for any given connector. The connector catalogue itself and the product overview PDF were not opened; under R25 they corroborate a grade that stands on the published count and the add-on line, and they are what would move this row. Verified 13 September 2026.

Deployment options are published with tenancy, regions and processing location. The security page states a single-tenant architecture with a separate AWS account, network, servers, databases and storage per client and dedicated resources per matter, deployment into any AWS region to meet local residency rules within forty-eight hours, hosting stated across the United States, European Union, United Kingdom, United Arab Emirates, Australia and Singapore, and the options of hosting in a client's own AWS environment or a government-approved FedRAMP environment; all processing and communication with cloud services occurs inside the client's private VPC over private links. The vendor contrasts this explicitly with shared multi-tenant platforms. Security page, security FAQ and partnership release read 6 September 2026.

Deployment is published in real detail and the tenancy model is stated, with region left general, which is a strong B under R38 where tenancy and region are co-equal limbs. Deployment options are not merely listed but sold as a choice: the purchase table names public cloud, private cloud and managed service, the product page adds on-premises and hybrid, and a deployment checklist is published to help a buyer choose between them. That range is wider than anything else in this lane and is a genuine answer for a buyer with sovereignty constraints, reinforced by a company-level sovereign cloud offering and a published private cloud availability map by geographic region. Tenancy is stated explicitly, which most records never manage: the Aviator FAQ confirms the feature set is available to OpenText eDiscovery cloud customers whether single or multi-tenant, so a buyer knows both that isolation is available and that shared tenancy is the alternative. The agreement names AWS as the cloud infrastructure for the eDiscovery Services and states that the AWS Customer Agreement governs the customer's use of it. What holds this off A is region. Availability is described as all AWS regions, which states reach rather than commitment: no region is named for a given customer, no residency guarantee is published, and nothing states where inference runs relative to where the collected evidence sits, though the FAQ's statement that the Bedrock session is created from the customer's own AWS environment implies they are co-located. Clause 10 adds that deletion after termination may be delayed up to 180 days by the infrastructure provider. Verified 13 September 2026.

Certification is real, dated and scoped, short of a report reachable without a sales conversation. The security page states a SOC 2 Type II audit completed in November 2025 by an independent third-party service auditor covering control environment and governance, risk assessment, logical access, system operations and change management, and monitoring and incident response; a penetration test completed in September 2025 by AppSecure, a CREST-accredited firm, with all findings remediated and verified; and a plan to pursue ISO 27001 in 2026, which is roadmap and not credited. The auditor is not named, and the SOC 2 report and pen-test summary are available to qualified prospects under NDA by contacting the team, which is the sales-gated tier. Security page read 6 September 2026.

Certification is real and stated, short of accessible evidence, which is the B band word for word. The standards are named and the assurance is real: the company describes an information security management system based on ISO/IEC 27001 incorporating the ISO 27002 control set and aligned to the NIST Cybersecurity Framework, states that it conducts regular Service Organization Controls audits, and publishes an information security paper naming SOC 2 with HITRUST, PCI DSS, HIPAA, SWIFT, TISAX, CyberEssentials+ and FedRAMP across its estate. The agreement turns that into an entitlement rather than a claim, clause 7.4 committing OpenText on written request to provide summaries of third-party audit reports and certifications applicable to the Services, naming SOC 1 Type II, SOC 2 Type II and the ISO 27001 certificate, and allowing a customer to verify scope or controls not covered by a report through a Shared Assessments Security Information Gathering form. Two things keep it from A, and both are about access and scope rather than substance. The route to evidence is gated: reports come on written request and only where the customer has entered a confidentiality agreement, which under R5 is the sales-adjacent tier rather than self-service, and no report, date or certificate is reachable without that step. And scope per product is not stated anywhere located, so a buyer cannot establish from the published material whether OpenText eDiscovery specifically sits inside the ISO or SOC boundary. The contrast with Docusign in this pull is instructive: the same class of vendor, the same standards, and an A there turned entirely on published scope and public registries. Verified 13 September 2026.

The vendor refers to multiple AI models without identifying them on the surfaces read. The Alchemy AI page is headlined three search engines and multiple AI models, the home page describes five AI engines, and a 2024 third-party directory entry refers to GPT-4, which is not credited; the vendor states that all processing occurs inside the client's locked-down AWS VPC, which places inference location with the client environment without naming the provider or model. No change-notification commitment is stated. The Alchemy AI page was not opened and is the rebuttal route. Home page, security page and navigation read 6 September 2026.

The supply chain is partly disclosed, the provider named and the architecture described without the model itself being named, which is the B band. The disclosure is among the most detailed located in this corpus and it sits in a product FAQ rather than a compliance document. The provider is named outright: Anthropic, reached through Amazon Bedrock via the Bedrock API, called directly from the customer's own OpenText eDiscovery AWS environment. Where it runs is stated, with availability in all AWS regions and the session initiated from the customer's own environment. The architecture is described in a way a buyer can actually reason about: Bedrock performs a deep copy of the model and hosts it, the provider has no read or write access and the model has no internet access, each interaction is a discrete project-scoped session, session data is deleted from Bedrock at close, and Bedrock does not store or log customer data in its service logs. The agreement supports it, naming AWS as an OpenText Licensor for the eDiscovery Services. Two A limbs fail and R34 keeps them separate from provider identification. The model itself is never named, no version or family is given, and nothing states what would happen if the underlying model changed. And no change notification commitment was located: the Data Processing Addendum's subprocessor mechanism gives a thirty-day objection and termination right on notice of a new subprocessor, which is a contractual route rather than a published commitment to notify on a model change. Verified 13 September 2026.

The unit and structure are stated without the figure on the surfaces read. Cloud Utility Pricing charges by the hour while a site is in use and a much lower rate, stated as seventy per cent less, when it is switched off, with scheduling of operating hours and an energy-saver power-down, and the vendor states most clients save fifty to sixty per cent on hosting; the 2023 launch describes a low hourly rate without naming it. A pricing page exists in the navigation and was not opened on 6 September 2026, and is the rebuttal route in either direction. Home page and 2023 launch release read 6 September 2026.

Pricing is gated behind a sales conversation while the capability set and the deployment split are published, so the shape is visible and the number is not, which is the C band. The shape is genuinely visible and better structured than most gated records. A How to buy table sets out what an eDiscovery subscription includes, naming OpenText eDiscovery Aviator, technology-assisted review, self-service data processing, predictive search and an integrated viewer with redaction, and identifies OpenText Connectors for Private Cloud as a paid add-on rather than bundled. A second table names the three commercial delivery options, being public cloud, private cloud and managed service, and a published deployment checklist helps a buyer choose between them. The Aviator page adds a third dimension, describing on-demand, multi-matter subscription and private cloud options for the AI features specifically. What is absent is any figure. No rate, unit, band, per-gigabyte or per-custodian price, minimum commitment or implementation charge appears anywhere located, and every route resolves to Contact us or Speak with an expert. One genuine piece of cost transparency exists inside the product and is recorded because it is unusual, though it is not published pricing: Aviator Review provides upfront token estimates and lets a team test criteria against sample data before committing to a full run, which the vendor markets as cost certainty. That tells a customer what a job will consume once they are already a customer, not what the platform costs. A pricing row is owed under R17 with no figure. Verified 13 September 2026.

Segment and coverage are described with substance; the boundaries are left open. The platform is sold to law firms, corporate legal departments and investigators for investigations, discovery and litigation, with the Alchemy page naming complex criminal matters, multi-defendant civil cases and joint defence groups as design targets, AI services extending to medical records, transcripts and scanned documents, review stated in any language, and hosting across six named jurisdictions. No matter type, data type or jurisdiction is named as unsupported. Home page, Alchemy page, security page and AI services navigation read 6 September 2026.

Coverage is described with real substance across buyer types and matter types, with the boundaries left open, which is the B band. Matter coverage is stated in the vendor's own framing rather than as a breadth claim: litigation, internal investigations, regulatory compliance, and personal data protection through automated detection and redaction driven by a pre-configured pattern library and custom regular expressions. Three use cases are published with detail behind each, covering case strategy and early case assessment, end-to-end collection through production, and personal data protection. Buyer coverage is evidenced rather than asserted, and it spans the three constituencies this product class serves: an investigating authority in the UK Serious Fraud Office, a corporate legal department in Bosch, and law firms in Gleiss Lutz, Eversheds Sutherland and Pillsbury, with Gleiss Lutz and the SFO showing the platform is sold outside the United States. A Legal industry solution page and a five-product Legal Tech family sit around it, and the training path is a four-day instructor-led case manager certification, which indicates the depth of configuration a real deployment involves. What is left open is the limit. No practice area or matter type is named as unsupported, no data volume ceiling is published, and no jurisdictional limit is stated. One boundary is stated and is credited to the vendor for candour, though it appears in the AI FAQ rather than on the coverage surfaces: validation of the Aviator features has been primarily on English-language documents and prompts. Verified 13 September 2026.