Luminance vs Robin AI: how they compare in 2026

The artificial intelligence is the product, and the vendor owns more of the stack than almost anyone on this index. Founded by mathematicians in 2015, it runs a multi model architecture it calls a Panel of Judges combining foundation, fine tuned and proprietary models, and ships its own legal intelligence model, Luna Crescent, trained in house and deployed in its own environment. Remove the models and there is no product.

The models are the product. Review, redlining, the Word add in's Ask, Draft, Edit and Research modes, clause level risk analysis and repository search are all generative or machine learning capabilities. The vendor's stated training base is more than 4.5 million legal documents and 100 million legal clauses. One qualification worth recording rather than discounting: the company also sells Managed Services combining the software with human legal review, so part of what it sells is people. The software product graded here is nonetheless model driven throughout.

Substantive disclosure with a real gap. The vendor publishes ContractIQ Bench, a proprietary benchmark of 189,000 manually annotated and reviewed data points assessing interpretation of named provision types including liability caps, termination for convenience and confidentiality obligations, tested on held out documents and concepts excluded from training, with blind evaluations by legal experts alongside. It publishes a result, 5 percent higher accuracy than leading general purpose models on contract understanding, and a speed figure of 200 to 400 tokens per second. It also states a design principle directly relevant to this axis: the model is trained to prioritise faithful extraction and to identify absence rather than invent an answer. Two gaps keep it off an A. The published figure is a relative delta with no absolute accuracy rate and no named comparator models, so a reader cannot tell what 5 percent higher is 5 percent higher than. And the benchmark is proprietary, with no sample tasks or rubric published, so an outsider cannot inspect or re run it.

Accuracy is asserted without measurement. Published grounding exists in outline: the Word add in has a Research mode described as enhancing answers with authoritative, up to date information from a curated set of trusted legal sources on the web, and review output is clause level with risk flags tied to company standards. What is not published is what makes this axis work: no accuracy figure, no hallucination rate, no test set, no evaluation methodology, and no description of how retrieval reaches those trusted sources or how a reader opens and verifies one. The headline claim across vendor material is a speed figure, 80 to 85 percent faster review, rather than an accuracy figure. Searched the site, the platform pages, the security page, Robin University and the guides and reports index on 29 Aug 2026.

A real published commitment with a described control mechanism, short of the full structure. The Panel of Judges architecture is itself an oversight design and is documented: multiple models analyse each clause independently and reach consensus, which the vendor states reduces hallucination risk. Outputs are described as traceable, and Traffic Light Analysis ranks deviation risk visually so a reviewer sees where to look. The vendor publishes a position piece arguing that human in the loop alone is insufficient and that systems must be designed for accuracy and transparency, which is a real stated philosophy rather than a slogan. Not located as of 29 Aug 2026: where the review point sits when the product negotiates with a counterparty directly, the threshold at which it escalates to a lawyer, and what the vendor commits to when an output is wrong. That first gap matters here more than for most, because the product sends agreements to counterparties and negotiates on the customer's behalf.

A real published commitment that the models work alongside a supervising lawyer, with documented review surfaces, short of the full control structure. The Word add in is described as surfacing clause level recommendations as tracked changes, comments or highlighted text based on company standards, which a lawyer accepts or rejects in place, and the four modes separate asking from drafting from editing so the user chooses the level of intervention. The company's Managed Services line puts its own legal professionals in the loop as a paid option, which is an unusually literal form of oversight. Not located as of 29 Aug 2026: any threshold at which the system defers, what it does on its own, and what the vendor commits to when an output is wrong.

Real deployment evidence with substance, short of attribution and method. Named customers appear in vendor and trade material including Hitachi, AMD, BBC Studios, Yokogawa and Koch, alongside a stated base of more than 700 organisations across 70 plus countries and all four of the Big Four consultancies. A customers page is published. The recurring figure, negotiation time reduced by up to 90 percent, is a vendor claim carrying a hedge and is not tied to any named customer, dated, or accompanied by a method. Searched the site, the customers page, the press releases and the resources index on 29 Aug 2026 and located no case study pairing a named organisation with figures and a date.

Customer logos and press mentions stand in for evidence, with results quoted from an internal source. A customers page exists and press logos including the Financial Times, Bloomberg, Forbes and CNBC are displayed prominently, alongside investor logos including Google, Temasek and PayPal. The most specific published figure, review completed over 85 percent faster, is attributed to the vendor's own internal legal team testing its own add in rather than to a customer. Searched the site, the customers page, the news and blog indexes and the guides and reports index on 29 Aug 2026 and located no named customer deployment carrying figures, a date and an assessable method. Press coverage and investor backing are not deployment evidence.

Substantive published commitments, short of the full picture. Segregation is the strongest element and is documented precisely: each customer receives a dedicated single tenant instance with complete isolation and no co mingling of data, which exceeds the level this buyer segment requires under the amended band. Access control is documented at an unusual depth, including that vendor staff cannot view customer documents without explicit authorisation given through the user interface, with all access tracked and audited. Encryption is specified to the key management service, cipher and rotation practice. Two gaps hold this off an A. No training prohibition on customer content was located anywhere in vendor material, which is a conspicuous absence for a vendor that trains its own models and publishes a corpus figure of 220 million legal documents. Attorney client privilege and work product handling is not addressed directly.

Confidentiality is asserted in general terms and the one commitment that matters carries a qualifier that changes it. The security page states that customer data will not be used for model training, fine tuning or other feature development without express consent. That is a consent gated permission rather than a prohibition, and it is the weakest training position located on this index so far. Real controls are published: AES-256 at rest, current TLS in transit, and a statement that data never leaves the vendor's AWS environment. Not located as of 29 Aug 2026: attorney client privilege or work product handling, any segregation model between customers, users or matters, and any retention or deletion terms. For a product whose core use case is ingesting counterparty contracts at volume, the absence of a segregation statement is material.

A boilerplate structure sits in the terms while the product is sold well beyond lawyers. Dedicated solution pages target compliance, executive, sales, procurement, finance, human resources and marketing teams alongside legal, and the product negotiates contracts on a customer's behalf. Searched the site, the solution pages, the published terms and conditions, the privacy policy and the resources index on 29 Aug 2026 and located no position on advice versus tooling, no treatment of competence or supervision duties, and no statement of jurisdiction limits, despite operation in more than 70 countries. This is the widest version of the non lawyer distribution question on the index so far, since the tool is marketed to marketing and HR departments.

The intended audience is lawyers and legal teams and the terms carry a standard structure, but no position on the advice line is published. Searched the site, the platform and services pages, the published terms and the privacy policy on 29 Aug 2026 and located no statement on advice versus tooling, no treatment of competence or supervision duties, and no jurisdiction limits, despite the company operating across the UK, US and Singapore and marketing to multiple jurisdictions. The Managed Services line, in which the vendor's own legal professionals perform review work for customers, raises the advice line question more sharply than a pure software product would, and nothing published addresses how that service is scoped or supervised.

Principles and architecture are published without a governance mechanism a buyer could audit. What exists is real and substantial: a published white paper on how the AI is built, a named Director of AI who authors technical material under his own name, a Cambridge based research team, a described validation regime through ContractIQ Bench, and a security advisory board of named external experts. But the security advisory board governs security rather than model behaviour, and no equivalent exists for AI governance. Not located as of 29 Aug 2026: a named owner of model governance, a pre release testing gate as distinct from benchmark results, an AI management certification such as ISO 42001, and anything on uneven output across matter types, parties or populations.

Searched the site, the security page, the trust centre entry point, the platform pages, the blog, the news index and the guides and reports index on 29 Aug 2026. No governance position for model behaviour was located: no named internal owner, no pre release testing regime, no published responsible AI or AI governance framework, no AI specific certification such as ISO 42001, and nothing on uneven output across matter types, parties or populations. The vendor publishes real security governance, which is a different subject. Rebuttable with a single link.

Substantive published policy covering most of the ground, at an unusual level of specificity. Published in the security FAQ: AWS Key Management Service encryption at S3 and EC2 level with AES-256 keys rotated regularly, TLS 1.2 or higher in transit, dedicated single tenant instances, role based and division level permissions configured by the customer under least privilege, configurable password and session timeout policy, mandatory staff security training, named threat detection through Darktrace's Enterprise Immune System and Juniper firewalls, and a described incident management process covering detection, mitigation and communication. Backups are stated precisely: nightly to a secondary AWS data centre in the same region, encrypted, retained a minimum of 14 days. Not located as of 29 Aug 2026: a retention period or deletion control for customer documents in normal operation as distinct from backups, and a named subprocessor list.

A generic security posture covers the product without addressing what happens to documents and prompts after processing. Published and real: AES-256 at rest, current TLS in transit, security integrated through the development lifecycle, data replication across geographic locations with disaster recovery exercises and automated failover, and named monitoring tooling in AWS CloudWatch, CloudTrail and Datadog. That is operational security described at a useful level of detail. What this axis asks for and did not locate as of 29 Aug 2026: any retention period for documents or prompts, any deletion control, a named subprocessor list, and an incident or breach notification practice. Anthropic and AWS are named as partners on the security page, which is a partial supply chain disclosure rather than a subprocessor list.

Liability is addressed only through published terms carrying a standard structure. Terms and conditions are published openly alongside a privacy policy, cookie policy and anti slavery statement, so a buyer can read the allocation of loss before entering a sales process, which keeps this above a pure absence. Searched those documents, the security page and the security standards white paper entry point on 29 Aug 2026 and located no indemnity running to the customer for third party claims arising from output, no warranty on output, no stated liability cap figure and no insurance position. Worth noting the product negotiates with counterparties on a customer's behalf, which raises the recourse question more sharply than a review only tool.

Liability is addressed only through published terms carrying a standard limitation structure. Terms and a privacy policy are published openly, so a buyer can read the allocation of loss before entering a sales process, which keeps this above a pure absence. Searched those documents, the security page and the trust centre entry point on 29 Aug 2026 and located no indemnity running to the customer for third party claims arising from output, no warranty on output, no stated liability cap figure and no insurance position. Worth noting that the Managed Services line involves the vendor performing legal review work, which raises a distinct recourse question that published material does not address.

Integrations are asserted at platform level with no documentation an implementer could use. The vendor describes connecting intake, negotiation, workflow and repository intelligence in one platform, and a Collaborate product exists for working with counterparties, and third party sources reference Microsoft Word and Outlook working surfaces. What was not located on the vendor's own property as of 29 Aug 2026, after checking the platform pages, the technology page, the security page and the resources index, is any integrations page, any named connector for document management, contract lifecycle, e signature, CRM or ERP, and any description of what an integration moves or what an administrator configures. For an enterprise platform sold to procurement and finance functions, that absence is notable.

Integrations are claimed without documentation an implementer could use. The Microsoft Word add in is real, documented at user level in Robin University, and is the primary surface, alongside a browser application with a searchable executed contract repository. Beyond Word, third party sources describe CRM and CLM integration but no vendor integrations page was located, and no named connector for document management, contract lifecycle or e signature was located on the property as of 29 Aug 2026. A third party review notes limited file format support beyond Word documents. Nothing published describes what any integration moves, in which direction, or what an administrator configures.

Deployment model is stated clearly with partial residency detail. Three things are published and specific: a dedicated single tenant AWS instance per customer with complete isolation, deployment within the customer's own environment as an alternative to the hosted option, and backup to a secondary AWS data centre within the same region, which confirms data stays in region. The vendor states AWS global infrastructure provides a solution tailored to geographic requirements. What is missing is the list: no named available regions, no statement of which regions a customer may select, and no statement of where processing happens as distinct from where data is stored. The single tenant and on premises options are genuinely stronger than most of this index.

Cloud delivery is stated and neither the tenancy model nor the region is stated. The vendor publishes that data never leaves its AWS environment and that data is replicated across multiple geographic locations for disaster recovery, which is a resilience statement rather than a residency one and arguably cuts against a customer wanting data confined to a jurisdiction. Not located as of 29 Aug 2026: any named region, any customer selectable residency, any tenancy model, and any statement of where processing happens as distinct from where data is stored. Third party sources mention private cloud deployment options, which is not vendor material and does not move this axis. For a UK founded vendor selling into the UK, US and Singapore, published residency options would be expected.

Certification is real and stated with correct nouns, short of accessible evidence. ISO 27001:2022 is named with its version and described as certification, and the SOC 2 Type 2 language is precise: the vendor says successful completion of a SOC 2 Type 2 examination assessing controls related to security, availability and confidentiality, which is the correct noun for SOC 2 and names the trust services criteria in scope. That is more careful phrasing than most vendors on this index manage. Regular independent third party penetration testing is stated. A named external security advisory board including a former Director General of MI5 and two former Darktrace executives is published with full biographies. What is missing is the evidence route: no trust portal was located, no report is downloadable or requestable through a published flow, and no coverage period, report date or auditor name was located as of 29 Aug 2026. A security standards white paper is published, which is the nearest thing to an evidence route.

Certification is real and stated with an open access route, short of scope and evidence. ISO 27001 and SOC 2 are both named as certified and GDPR compliance is stated, and a trust centre is published at a stable URL as a self serve route rather than a sales gate, which under the three tier test is materially better than absent. What was not located as of 29 Aug 2026 is any coverage period, audit scope, report date, SOC 2 type designation or named auditing firm. Note the vendor writes ISO without a number in one place on the security page and ISO27001 in another; the certification is stated clearly enough elsewhere that this reads as loose copy rather than a claim problem, and it is recorded rather than concluded on.

The supply chain is partly disclosed and the architecture is described in more depth than most. The vendor publishes that it runs a multi model Panel of Judges combining foundation, fine tuned and proprietary models, that it continuously evaluates and selects the best model per task, and that its own model Luna Crescent is deployed inside its own AWS environment. It states the supply chain consequence explicitly and in the customer's terms: owning the model reduces reliance on external model providers, limits data exposure to additional subprocessors, and protects customers from third party availability, pricing and access disruption. Hosting is named as AWS. What is not published is which foundation models sit in the panel, from which providers, or which tasks route to them, and no subprocessor list or change notification commitment was located as of 29 Aug 2026. A buyer therefore knows the shape of the chain and its own model, but not the third party links in it.

The vendor identifies a partner without disclosing the supply chain. Anthropic and AWS are both named as partners on the security page, and third party coverage describes collaboration with Anthropic and AWS to build secure models, so a reader can infer the likely model provider. But naming a partnership is not the same as stating which model serves which task. Not located as of 29 Aug 2026: which models are used in production, where they run, whether any other provider is involved, a subprocessor list, or any commitment to notify customers when the supply chain changes. Graded on what is published rather than on the inference the partner logo invites.

Checked the site navigation, the platform and solution pages, the customers page, the about section and the footer on 29 Aug 2026. No pricing page exists on the property, no rate is published, no unit of charge is stated and no tier structure appears. Every commercial path terminates in a demo request. No free trial or self serve entry point was located, and no third party pricing figure was located either.

Checked the site navigation, the platform and services pages, the customers page and the footer on 29 Aug 2026. No pricing page exists on the property, no rate is published, no unit of charge is stated and no tier structure appears. Every commercial path terminates in a demo request. Vendor material references a free account with add in access and paid accounts with playbook customisation, and directs the reader to speak to a member of the team to understand what paid accounts include, so even the feature split between free and paid is not published. Third party sources describe subscription pricing without figures.

Who the product serves is documented precisely across two dimensions, each with its own published pages. Six industries: manufacturing, financial services, pharmaceutical, technology, insurance and chemical. Eight business functions: legal, compliance, executive, sales, procurement, finance, human resources and marketing. Both law firms and corporate legal departments are addressed, with a stated base of more than 700 organisations across more than 70 countries including all four Big Four consultancies and named enterprises. An academic programme is published as a separate segment. The practice boundary is clear from the structure and consistent throughout: this is contract work end to end, from generation through negotiation to post execution analysis and investigation, and nothing on the property claims litigation or research capability it does not have.

Coverage is claimed broadly without evidence that the breadth is real. Vendor material addresses in house legal teams, law firms and corporate clients across multiple jurisdictions, and the practice focus is clearly contract work: review, drafting, negotiation and post signature obligation tracking, with third party sources naming NDAs, service agreements, procurement and employment agreements. Three offices are published across London, New York and Singapore. What was not located as of 29 Aug 2026 is any segmentation a buyer could use: no pages by firm size, industry or role, no statement of which practice areas or team sizes the product is built for, and no statement of what it is not built for. The customers page exists but does not resolve the segments claimed.

Searched the security page and its FAQ, the technology page, the Luna Crescent announcement, the published terms and conditions, the privacy policy and the resources index on 29 Aug 2026. No located term or policy states whether customer content may be used to train models, either way. The absence is more consequential here than for a vendor that only calls third party APIs: this vendor trains its own models, publishes a corpus figure of more than 220 million verified legal documents accumulated over a decade of platform use, and describes that corpus as something general purpose providers cannot replicate. Vendor material states the platform has been exposed to those documents without stating whether customer documents are among them or on what basis. Recorded as silent under the rule that a value is never inferred from the absence of a contradiction.

The security page states that customer data will not be used for model training, fine tuning or other feature development without express consent. Read precisely, that is a commitment not to train absent consent rather than a prohibition, so training becomes available where consent is given, and no material was located describing how consent is sought, at what level it is given, or whether it is a contract term or a product setting. The vendor's trust centre separately states data is never used to train other AI models, which addresses third party models rather than its own. Recorded at the opt in value as the closest published fit, with the qualifier and the ambiguity stated here rather than resolved.

Retention is acknowledged and partly quantified without a period for the primary system. The security FAQ states each customer instance is backed up nightly to a secondary AWS data centre in the same region, encrypted, and kept for a minimum of 14 days, which is a real published figure and a floor a buyer can plan against. What was not located as of 29 Aug 2026 is any retention period for documents, prompts or outputs in normal operation, any customer control over that window, or any deletion commitment. A minimum backup retention states how long data persists after deletion rather than how long it is held.

Searched the security page, the trust centre entry point, the published terms, the privacy policy and the help centre entry point on 29 Aug 2026. No public material states how long documents, prompts or outputs are retained, whether a customer controls the window, or whether deletion is available. The vendor does publish that data never leaves its AWS environment and that data is replicated across multiple geographic locations for disaster recovery, which describes where copies sit rather than for how long. The browser application maintains a searchable repository of executed contracts, so the product is designed to retain documents, with no published terms attached.

The product maintains its own documented permission model and documents it more thoroughly than any other record on this index. Between customers, isolation is architectural: a dedicated single tenant instance each, stated to ensure complete isolation with no co mingling of data. Inside a customer, the vendor publishes division level permissions administered by the customer under least privilege, role based configuration, customer configurable password and session timeout policy, and a statement that vendor staff cannot view customer documents without explicit authorisation given through the interface, with all access tracked and audited. Division level permissions are the nearest published equivalent to a wall. Recorded at the own model value rather than the positive one because no document management integration was located whose permissions retrieval could inherit at query time, and because conflicts and ethical walls are not addressed as such.

Searched the security page, the trust centre entry point, the platform pages, the published terms and the help centre entry point on 29 Aug 2026. No vendor material addresses segregation of any kind: not between customers, not between users, not between matters. Third party sources refer to sophisticated user permissions and single sign on, which is not vendor material and was not treated as evidence. No document management integration was located whose permissions retrieval could inherit. The product ingests counterparty contracts at volume and maintains a shared searchable repository, which makes the absence of any published segregation model a live question rather than an academic one.

Searched the published terms and conditions, the privacy policy, the security page and its FAQ, and the security standards white paper entry point on 29 Aug 2026. 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. Noted for a future reader: the single tenant and on premises deployment options materially change what a vendor could produce in response to such a request, but the vendor does not make that argument in published material and it is not recorded as a value here.

Searched the published terms, the privacy policy, the security page and the trust centre entry point on 29 Aug 2026. 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. This records a search across the published documents that did not surface the clause rather than a reading of every document end to end.

The corpus here is contract and legal document data underpinning a proprietary model rather than primary law, and it is quantified in detail without being sourced. Published: more than 220 million verified legal documents the platform has been exposed to over a decade, a curated training selection spanning roughly 3.4 million legal concepts and data points, breadth described as spanning virtually every industry including complex agreements, difficult file formats and obscure drafting styles, and a separate benchmark corpus of 189,000 manually annotated data points. What is not published is where any of it came from or on what rights basis it was assembled. Trade coverage notes that many documents in the training base were not publicly disclosed. Recorded at the weakest value because scale and character are described while the source and licence basis are not.

Two corpora are described and neither is identified. The training base is quantified as more than 4.5 million legal documents and 100 million legal clauses, with no statement of where those documents came from or on what rights basis they were assembled, which is the question this signal exists to ask and matters more for a corpus of contracts than for public case law. Separately the Word add in's Research mode is described as drawing on a curated set of trusted legal sources on the web, with none of those sources named. No licence basis, jurisdiction list or update cadence was located for either as of 29 Aug 2026.

Searched the site, the six platform product pages, the technology page 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 a contract lifecycle product whose grounding is contract language and the customer's own precedent rather than case law, so a citator is largely outside its design.

Searched the site, the platform pages, Robin University and the help centre entry point 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 a contract review product whose grounding is customer playbooks and contract language rather than case law, so a citator is largely outside its design, though the add in's Research mode does reach legal sources on the web.

The vendor documents abstention behaviour as an explicit design objective, which is the first time this signal has recorded anything above an absence on this index. Published: the proprietary model is trained to prioritise faithful extraction from source documents, and where information is not present it is designed to identify that absence rather than invent an answer, with the vendor stating this is what instils trust for legal professionals. Vendor material separately describes flagging what is absent as well as what is present as a capability advantage. Recorded at the documented value rather than the demonstrable one because no published evaluation of the abstention behaviour itself was located as of 29 Aug 2026: the ContractIQ Bench results address interpretation accuracy rather than refusal rate.

Searched the site, the platform pages, Robin University, the blog and the help centre entry point on 29 Aug 2026. No published material describes what the product does when it cannot ground an answer, and no explicit no answer path was located. Review output carries risk flags at clause level, which ranks how serious a flagged issue is rather than how confident the system is that it found one, and was not treated as a confidence signal for this purpose.

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 that this is a contract lifecycle product rather than a litigation or research tool, so its output is unlikely to reach a court filing as cited authority, and note that the database is weighted toward US filings while this vendor is UK founded and operates across 70 plus countries.

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 that this is a contract review product rather than a litigation or research tool, so its output is unlikely to reach a court filing as cited authority, and note that the database is weighted toward US filings while this vendor is UK founded.

Searched the site, the insights and white papers indexes, the press releases and the resources hub on 29 Aug 2026. No engagement with any named ethics opinion or professional guidance was located, including ABA Formal Opinion 512, US state bar guidance, and Solicitors Regulation Authority or Law Society guidance given the company's UK base. The vendor publishes substantial thought leadership on AI reliability, including a piece arguing that human in the loop alone is insufficient, which engages with the professional risk question in substance while naming no guidance a buyer is bound by.

Searched the site, the blog, the news index, Robin University, the guides and reports index and the webinars page on 29 Aug 2026. No engagement with any named ethics opinion or professional guidance was located, including ABA Formal Opinion 512, US state bar guidance, and Solicitors Regulation Authority or Law Society guidance given the company's London base and UK founding. The vendor publishes substantial educational content about contract workflows and legal AI adoption, which addresses practice efficiency rather than the professional responsibility obligations its buyers are bound by.

Savings are claimed with nothing published on the client's side of the equation. The recurring published claim is negotiation time reduced by up to 90 percent, alongside speed framing throughout including a stated four times faster generation than generalist tools. Searched the site, the platform pages, the customers page 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. The vendor sells to law firms as well as corporate teams, so the firm side of that question applies.

Savings are claimed prominently with nothing published on the client's side of the equation. The headline claim across vendor material is contract review 80 percent faster, with an internal figure of over 85 percent faster from the vendor's own legal team. Searched the site, the platform and services pages, the blog and the guides and reports 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. The vendor sells to law firms as well as in house teams, so the firm side of that question applies here.

Substantial security material is published openly, including a detailed security FAQ, named certifications, a named external security advisory board and a security standards white paper, all reachable without a sales conversation. But the specific artifacts this signal turns on were not located as of 29 Aug 2026: no subprocessor list, no statement of which model providers see customer content, and no client facing consent or notification pack a firm could forward to its own client. The vendor does state that owning its model limits data exposure to additional subprocessors, which is an argument about the shape of the chain rather than a disclosure of it. Recorded as not addressed because no list exists to point to.

A trust centre is published at a stable URL and is reachable without a sales conversation, which is a real access route, and the security page names Anthropic and AWS as partners. But the artifacts this signal turns on were not located as of 29 Aug 2026: no subprocessor list, no statement of which model providers see customer content, and no client facing consent or notification material a firm could forward to its own client. A partner logo is not a model provider disclosure. Recorded as not addressed rather than at the subprocessor value because no list exists to point to.

Searched the site, the platform product pages, the technology page and the security page on 29 Aug 2026. Vendor material states outputs are traceable and that all data access is tracked and audited, so elements of an access trail exist. But no per document export covering model used, sources retrieved and human verification together was located, and the model used would be difficult to state in any case given the Panel of Judges architecture routes tasks across multiple models. Noted for context: this is a contracting product rather than a litigation product, so a judicial AI disclosure order is less likely to reach its output.

Searched the site, the platform pages, the security page and Robin University on 29 Aug 2026. Third party sources describe exportable audit logs and reports, which is not vendor material and was not treated as evidence. On the vendor's own property no per document export covering model used, sources retrieved and human verification together was located, and the model used is not identifiable from published material in any case. Noted for context: this is a contracting product rather than a litigation product, so a judicial AI disclosure order is less likely to reach its output.