Gaius-Lex vs Omnilex: how they compare in 2026
Gaius-Lex and Omnilex both sell AI research and drafting grounded in national law, but for different countries: Gaius-Lex covers Poland, the Czech Republic, Germany and Austria, and Omnilex covers Switzerland. Omnilex sits in the top two bands on thirteen of fifteen axes and Gaius-Lex on nine of fifteen, identical on eight. Omnilex leads on security, evidence and professional secrecy. It holds ISO 27001 with a trust center carrying a dated penetration test, names customers including the Swiss Confederation and Bayer, and engages Swiss professional secrecy and auxiliary person liability by name. Gaius-Lex's lead is deployment choice and a named model supply chain. It offers SaaS, hybrid or on premise installation and names OpenAI, Anthropic and three other providers behind its AI. Its terms reserve the right to use query history to develop its AI, while its English site says client files never train any model. Omnilex does not say whether customer content trains models.
At a glance
All 15 axes, side by side
The same grid applied to every vendor in the index, graded from public sources. Hover a grade to see what the letter means on that axis.
AI Centrality
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
The agent is the product. The terms of service define the paid services as putting questions to a language model and issuing instructions to it, and the metered unit of charge, called a Lex, is consumed per AI action. There is a classic text search alongside it, described in the terms as a separate service, but every function the product is sold on runs through the model.
The models are the engine of what a buyer uses and a substantial content asset sits beneath them. The corpus is counted openly at 1,057,541 sources across five categories, dominated by 797,321 Swiss court decisions and 22,181 laws and articles, and that collection would remain a usable Swiss law database if every model were removed, which is the product several incumbents in this market already sell. What the AI adds is the reasoning layer over it, plus one thing that is genuinely model-created: the Omnilex Commentary, 222,525 annotations the company writes itself article by article on Swiss law, which exists only because the models produced it. Agents built without code from a firm's own playbooks sit on top. The pricing page frames the whole offering as one license covering corpus, models and agents together, which is an accurate description of a product where the machine learning is the mechanism and the collection is the substrate. Verified 2 September 2026.
Citation Accuracy and Hallucination Disclosure
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
Grounding is documented in unusual detail. Every thesis carries a footnote and a link to the regulation or ruling, case signatures are checked for existence against the database, and the vendor states that where confirmation is absent the system flags it rather than asserting. No accuracy rate, hallucination rate or published evaluation exists, which holds the row at B. Recorded alongside that, because the two sit in the same record: the marketing estate carries the absolute claims that the product never guesses and produces no hallucinations, while the terms of service state that the AI does not always answer in accordance with reality, name the phenomenon as hallucinating, and warn that it may materially affect the result. The contractual limitation disclosure is genuinely candid and is credited here. The marketing absolute above it is not evidence.
Grounding is real, documented and unusually well labeled. The stated design is that every claim is cited and ready for the file, and the Omnilex Commentary carries the line that every statement is citable, with the worked example on the home page showing an annotation to article 336c of the Code of Obligations footnoted back to its authorities. The user also selects which source types the research draws on, choosing between court rulings, commentary or internal data, which is a described retrieval control rather than an opaque one. One practice deserves specific credit and is rare in this corpus: the commentary the vendor generates is expressly labeled as AI-drafted, so a lawyer can tell machine-written analysis from published doctrine. What is absent is measurement. No accuracy figure, evaluation, benchmark or test set appears anywhere. Clause 6 of the terms goes the other way and disclaims it directly, stating that the provider accepts no liability for the accuracy or legal usability of generated content, which is candour about the limitation rather than evidence about the rate. Verified 2 September 2026.
Autonomy and Oversight Model
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
The product is marketed at the high autonomy end, as an autonomous trainee that plans its own research and delivers drafts, and the vendor sets against that a contractual statement that generated content is never self standing, is to be treated only as a suggested formulation, and requires the user's own work at the concept, creation and review stages. A contractual constraint is materially stronger than a marketing assurance of human review, which is why this is a high B. It stays at B because it is a general statement rather than one attached to a named plan or mode stating what that tier's output may not be used for.
The review obligation is contractual and the agent layer is user-triggered. Clause 6 of the terms states that the user remains responsible for reviewing, adapting and using all content, which places the checkpoint on the lawyer in the operative document rather than in marketing. The product framing is consistent with it, describing AI-assisted suggestions at clause 2 and, on the product page, a system that analyses documents, laws and judgments, structures arguments and highlights relevant content. Agents are described as built from a firm's own playbooks and review steps and run on demand, so they execute when a person starts them rather than autonomously, and the worked example sets out its four steps openly. What is not published is the boundary. Nothing states what an agent completes without a human once started, what happens when a step fails or returns nothing, or what threshold separates a suggestion from an action. The firm authors the workflow, so the operating envelope is customer-set and undescribed. Verified 2 September 2026.
Operational and Outcome Evidence
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
Checked 15 September 2026 across the English and Polish estates, including the mission, team, blog, guides and webinar surfaces. No named customer, no case study and no production figure of any kind is published by the vendor. There is substantial Polish trade press coverage and an investor blog post referencing dozens of early clients, but both are third party and the investor is not a neutral source. The product is plainly operating, with a live checkout, an application and four jurisdictions, and operating is not outcome evidence.
Ten customers are named by logo across private practice, insurance, corporates and the state: the Swiss Confederation, Bayer, the legal-protection insurer Dextra, the pension fund Asga, the disability charity Pro Infirmis, and the firms SwissLegal, Ruoss Voegele, Neuwerk, Domenig and Seeger Frick. Two carry dedicated case pages with named individuals speaking on video, Florian Probala, Chief Claims Officer at Dextra, and Dr. iur. Marco S. Marty, Founding Partner at MV Legal Partners. A named public authority among the references is worth noting on its own, because a federal administration adopting an AI research tool is a procurement signal a buyer can weigh. One figure is published, that over 6,000 lawyers use the platform, described as the most-used AI workspace in Swiss law. What holds this below A is measurement: the 6,000 figure is an adoption count rather than an outcome, no figure is attached to any named customer on the pages read, and nothing is dated. The two case pages were not opened this pass. Verified 2 September 2026.
Privilege and Confidentiality Posture
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
Real controls exist: encryption in transit and at rest, automatic or manual anonymization per document, and a contractual bar on the user submitting third party personal data at all unless a data processing agreement has been signed, with an obligation to anonymize every document until it is. Two things hold the row at C rather than B. First, neither privilege, work product nor professional secrecy is treated anywhere as a duty the vendor owes, and no matter level segregation is described. Second, the confidentiality picture in the agreement runs against the marketing: the terms reserve a right to use query history to develop the vendor's AI algorithms, while the English estate states that client files, cases and analyses are never used to train any model by the vendor or any third party. The signal rows carry the detail.
The privilege limb is met squarely, which most records in this pull fail, and the training limb is not addressed at all. Omnilex engages the professional-secrecy question in its own legal system's terms rather than gesturing at confidentiality: the home page states that models are processed on Swiss infrastructure so that attorney-client privilege stays intact, that the platform was built for professional secrecy from day one, and that the license agreement with its data processing agreement is drafted to respect professional secrecy and auxiliary-person liability. That last concept is the Swiss doctrine governing a lawyer's responsibility for the third parties they engage, and naming it is a substantive engagement with the buyer's own obligations. Retention and deletion are addressed at clause 7 of the terms, with data deleted after the contract ends subject to statutory retention. The position on model providers is partly answered by keeping processing on Swiss infrastructure. Two limbs fail. Nothing published describes segregation between customers, users or matters, despite the product organizing work into named projects and mandates. And nothing located addresses whether customer content is used for training, in either direction, as set out on that signal. Verified 2 September 2026.
UPL and Professional Responsibility Posture
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
The strongest posture located on this axis in the corpus so far, and it is contractual rather than marketing. The terms state in terms that the provider does not supply legal assistance, that an answer does not constitute legal assistance, and that a user taking any legal step should consult a person professionally qualified to give it. Eligibility is restricted to entrepreneurs contracting in direct connection with their business or profession, consumers are routed to a separate individually negotiated arrangement through a dedicated address, and a separate student agreement grants an educational license that expressly bars commercial, paid or professional use including during a law firm internship. What holds it below A is that no bar or ethics guidance is named anywhere, and that the terms address Polish law while the product is now marketed across four jurisdictions, so the jurisdictional limits of the posture are not stated.
A real position on tooling versus advice is published in the operative terms rather than a footer. Clause 2 characterizes the service as software that supports users in working on legal questions with AI-assisted suggestions, which frames output as input to a lawyer's work. Clause 6 states that the provider accepts no liability for the accuracy or legal usability of generated content and that the user remains responsible for reviewing, adapting and using all content, which is the supervision dimension stated as an obligation. Coverage is bounded in practice, since the product, the corpus and the commentary are Swiss law throughout and the terms are governed by Swiss law with jurisdiction in Zurich. The professional frame is engaged more directly than usual through the references to professional secrecy and auxiliary-person liability. Two things are missing. No express statement that output is not legal advice, and no statement about the professional relationship, appears anywhere on the surfaces read. And no bar or professional body guidance is named, with neither the BGFA nor Swiss Bar Association material referenced, which is recorded on the bar guidance signal. Verified 2 September 2026.
AI Governance and Bias Disclosure
Published governance over model behavior: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
Checked 15 September 2026 across both estates, the terms of service, the student terms and the privacy policy. No governance owner, no pre release testing description, no evaluation practice and no statement about output variation across matter types, jurisdictions or parties. The terms do disclose that the model is developed on limited data sets whose content the provider does not answer for, which is a limitation disclosure and is credited on Citation Accuracy rather than here.
A governance surface exists and its contents could not be read. The SafeBase trust center carries a dedicated AI section listing AI Security, AI Governance and AI Monitoring as published items, which is more structure than most records in this corpus offer and is evidence that a position exists rather than that none does. The item pages themselves are driven by query parameters that did not render, so what those headings contain is unestablished, and this grade rests on the existence of the surface rather than on any substance credited from it. Around it sit an ISO/IEC 27001:2022 certification audited across the whole organization, a dated penetration test report, and an ESG section covering anti-bribery, anti-competitive practices and modern slavery, all of which are adjacent subjects this axis treats separately. What was located nowhere, on any surface, is anything about bias or uneven output: no statement addresses whether the system performs differently across cantons, languages, court levels or party types, which matters for a product whose corpus spans a federal system with three official languages. Verified 2 September 2026.
AI Safety and Data Stewardship
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
The subprocessor position is the strongest element and is rare in this corpus: the privacy policy names the providers whose engines power the AI, being OpenAI, OctoML, Gladia, Anthropic and RunPod, and states that each processes under a data processing agreement using European Commission standard contractual clauses. The list has demonstrably been maintained, since an earlier published version named only two of the five. Encryption at rest and in transit, per document anonymization, account deletion within 24 hours of request, and an export of stored query history on account blocking are all published. Held at B by the absence of any stated retention period for documents or prompts and of any incident response practice.
Substantive across most of the ground, short of specificity on two elements. Access control is the strongest part and is stated as the default set legal teams expect: SAML single sign-on, audit logs, IP allow-listing and data lifecycle management, with the trust center adding credential management, brute-force protection and account hardening, endpoint detection and response and disk encryption. Encryption is stated at rest and in transit. Subprocessors are named, with the trust center legal section identifying Amazon Web Services and Microsoft Azure. Incident practice is stated, with a dedicated response team and an incident response item in the trust center, and backups, business continuity and disaster recovery and anti-DDoS all appear as published items. What keeps this off the top band is that two of the five elements are stated without being specific enough to hold the vendor to. Retention is given only as data deleted after the end of the contract unless statutory retention applies, with no period for anything during the term. And the incident commitment carries no notification deadline to the customer, the trust center saying instead that more detail is available on request. The named subprocessors are also infrastructure only, which is examined on the model supply chain axis. Verified 2 September 2026.
AI Liability and Recourse
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
The position is published, specific and entirely negative, which is worth more to a buyer than silence and is why this sits above D. The terms exclude the provider's liability for the quality of AI output and set out six stated reasons, including that the system depends on third party providers running the AI engine, that the model is built on limited data sets, and that it may hallucinate. No indemnity, no liability cap, no carve out and no insurance position is published. The terms also record that an individually negotiated agreement takes precedence where it conflicts, so an enterprise buyer may be able to negotiate something the published document does not offer.
A real position is published and it is short of the full picture. Clause 6 of the terms sets out the allocation in four sentences: use is at the user's own risk; the provider accepts no liability for the accuracy or legal usability of generated content; the user remains responsible for reviewing, adapting and using all content; and the provider's liability is limited to gross negligence and intent. That last limb is more informative than it looks, because it is the maximum exclusion Swiss law permits, so a reader knows precisely where they stand, with nothing recoverable for ordinary negligence and no monetary cap at all on the residual liability for gross negligence and intent. Cyber insurance is separately published as a document in the trust center, which is one of the four things the top band asks for. What is absent is the rest. No indemnity runs in either direction, no warranty is given on output and clause 6 expressly disclaims it, and no cap figure appears. The availability commitment at clause 5 is weak and honest about it, targeting at least 90 percent outside maintenance windows and stating it is not guaranteed. One scope limit belongs on the record: clause 9 states that use is based on a separate license agreement, and that agreement is not published, sitting instead as a template behind the trust center's access flow, so what a signed customer actually receives is not established. Verified 2 September 2026.
Practice Systems Integration Depth
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
Checked 15 September 2026 across both estates, the feature pages and the terms of service. No document management system integration is published, and iManage, NetDocuments, SharePoint and OneDrive appear nowhere. There is no Word add in and no Outlook add in. The product is a browser application, defined in the terms as software run in a web browser, into which a user uploads files in listed formats. Documents live in the vendor's own store rather than in the firm's. This is the widest gap on the record against every firm facing competitor in the lane.
Integration is claimed and never specified. The product page states that a firm can connect its document management and legal tools without media breaks and keep its proven workflows, which is a direct claim to integrate with the systems legal work already lives in. Not one of those systems is named. No document management product is identified, no practice management or e-filing tool, no connector list, and no integrations page exists. Enterprise plans are stated to include custom integrations and APIs, but that is a scoping conversation rather than a published capability, and no API or developer documentation was located. What is concrete is authentication rather than workflow: SAML single sign-on across plans and company SSO at enterprise level. A claim to integrate, with the systems unnamed and nothing an implementer could use, is what this band describes. Checked the home page, product page, pricing page, terms, privacy policy and trust center on 2 September 2026. Verified 2 September 2026.
Deployment Model and Data Residency
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Deployment is genuinely strong: SaaS, hybrid and on premise are all published as buyer choices rather than enterprise conversations, and running locally is the core pitch of the product. Residency is where it comes apart, and the row is graded on both. The English estate states that every uploaded document is processed exclusively on servers within the European Economic Area and never routed outside it. The privacy policy names US incorporated providers among those receiving user queries, and an earlier published version of the same policy expressly contemplated transfers outside the European Economic Area under standard contractual clauses. The on premise product may well resolve this, since nothing leaves the customer estate, but no surface states that the residency claim is scoped to one deployment shape. Published options earn the B; an unresolved residency claim prevents the A.
Residency is stated with a named city and a named cloud, and the tenancy limb is silent. Omnilex publishes that it is hosted on Microsoft Azure in Zurich, that data is encrypted at rest and in transit and hosted entirely on its Swiss Azure cloud, and that models run on Swiss infrastructure, which is the point on which the privilege claim rests. Processing is separated from hosting and bounded, with the pricing FAQ stating that data is processed in Switzerland and in selected EU countries recognized as adequate by the Swiss Federal Data Protection and Information Commissioner. Naming the supervisory authority rather than asserting adequacy is a good habit. Two things hold this below the top band. No tenancy or isolation model is published anywhere: nothing states whether the platform is multi-tenant, single-tenant or how one firm's data is separated from another's, which is the limb this axis pairs with region. And there is a small tension a buyer should notice between hosting stated as entirely on the Swiss cloud and processing stated as extending to selected EU countries; both are published, neither is reconciled. No on-premises or customer-cloud option is offered. Verified 2 September 2026.
Security Certifications and Trust Center
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
Checked 15 September 2026, including the Polish security page, both estates, the terms of service and the privacy policy. No SOC 2, no ISO 27001, no penetration test summary, no trust portal and no attestation of any kind, claimed or reachable, and no self serve request route. The European Union funding badge in the footer is a grant marque and not a certification. The architecture the vendor does publish is credited on AI Safety and Data Stewardship.
A real trust center exists, it renders, and it is inventoried. The SafeBase portal at trust.omnilex.ai lists ISO/IEC 27001:2022 under compliance, and among its documents a penetration test report dated 5 September 2025, the Omnilex License Agreement including a DPA template, technical documentation, cyber insurance, a data security policy and a network diagram, alongside sections for product, data, application and endpoint security, infrastructure and legal. The certification is described on the home page as audited across the whole organization, which is a scope statement rather than a bare badge, and a document carrying a date is visible without any request. What is not established is the access tier for the documents themselves. The portal offers a public and private split with a get-access flow, and it does not state whether access is granted instantly on an email or an NDA click-through, or only after review, so the lower tier is graded and the reason recorded. No auditor or certification body is named, and no certificate number, issue date or validity period is published. No request was submitted. Verified 2 September 2026.
Model Supply Chain Disclosure
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
Among the better disclosures in the corpus, and it sits in the privacy policy rather than in marketing. The vendor names five providers supplying the engines behind its AI: OpenAI, OctoML, Gladia, Anthropic and RunPod, with the transfer basis stated as standard contractual clauses. Comparing published versions shows the list being maintained as the stack changed. Two things hold it at B. No specific model or model version is named, only the providers. And the disclosure is in tension with the marketing, which positions the product as running locally deployed small language models without connecting to public cloud AI services; naming OpenAI and Anthropic as recipients of user queries is a different architecture from the one the English estate describes. No commitment to notify customers when the list changes is located.
The models are named to the version on the home page, which nothing else in this pull does, and the disclosure stops short of the rest. Five are listed as selectable, GPT-5.6, Claude Opus 5, Mistral Large 3, Kimi K2.6 and Grok 4.6, with the user choosing which answers a given question, and the accompanying claim is that every frontier model is available and processed on Swiss infrastructure so that attorney-client privilege stays intact. Where inference runs is therefore stated at estate level, on Microsoft Azure in Zurich. The trust center names Amazon Web Services and Microsoft Azure as subprocessors. Two limbs are missing. No model house is identified as a legal entity and none appears in the subprocessor list, so a reader can see that Claude Opus 5 is available without learning which entity contracts for it or what it may retain, and the inference that the models run inside the Swiss cloud estate rather than through the vendors' own APIs is left to be drawn rather than stated. And no commitment to notify customers when the model set changes was located, which matters on a product that markets access to whichever frontier model is current. Verified 2 September 2026.
Commercial Transparency
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
A buyer can price this without speaking to anyone. Two plans are published with rates, Gaius Standard at 159 euro per month net and Gaius Pro from 2,150 euro per month net, each with a live checkout, and a seven day full trial requires no card. The unit of charge is published and defined at length in the terms: a Lex is an internal accounting unit consumed per action, plans carry a stated Lex allowance, a consumption table sets the cost of each action, and the rules on top ups, carry over, freezing on non renewal and expiry after 90 days are all written down. Team plans pool Lexes across seats with each additional seat priced. Billing periods of 30, 90 and 365 days are offered and price changes require 14 days notice with a right to cancel. A free tier and a student plan sit below the paid rate.
The strongest commercial disclosure located in this pull. Rates are published with the unit and the term: CHF 2,000 per user billed annually for a single user, CHF 1,600 per user billed annually for teams from two users, and a custom tailored agreement for enterprise from twenty users, all stated excluding VAT, with an annual and monthly toggle so a buyer can compare billing cadences. The unit is explicit and the vendor makes a point of it, stating that the price scales only with the people who use it and that corpus, models, agents and security are in every plan. The limb this axis asks for that almost no vendor answers is answered directly: implementation is not an extra. Onboarding, set-up, training and support are listed as included in every plan without asterisks, and the FAQ states that nothing essential is held back in a higher tier, with Team adding collaboration and know-how integration and Enterprise adding SSO, custom integrations and APIs. A fourteen day free trial of the full product requires no card and no commitment. Usage limits are published too, at up to 1,000 pages and 500MB per document with otherwise unlimited use. The only withheld figure is the enterprise tier, which the page explains is a scoping conversation about integrations and support rather than a different product. Verified 2 September 2026.
Firm and Practice Coverage
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
Eight buyer segments are named with a distinct stated use for each: law firms, solo practitioners, tax advisers, courts, public administration, financial institutions, corporations and small and medium business, with a student plan alongside. Jurisdictional coverage is specific rather than claimed, naming the primary sources per jurisdiction across Poland, the Czech Republic, Germany and Austria. What holds it below A is firm size, which is addressed only through the two plan tiers and a team seat model, and the absence of any practice area breakdown beyond tax being called out separately.
Four buyer segments each have their own page and each is described in terms of the work rather than as a label: law firms handling research, drafting and review across mandates; in-house teams moving from contract review to regulatory questions; public administrations applying federal and cantonal sources; and insurers assessing claims against statute and precedent. Covering government and insurance explicitly is unusual and is borne out by the named references, which include the Swiss Confederation and a legal-protection insurer. Jurisdictional coverage is stated precisely through the corpus itself, with federal and cantonal decisions, statutes, official documents and literature counted by category. What is missing is the boundary. No practice areas are enumerated, and while the worked examples on the home page range across employment, tenancy and accident insurance law, that is illustration rather than a coverage statement. No firm size band is given beyond the seat thresholds in pricing, nothing addresses whether the product serves any jurisdiction other than Switzerland, and nothing states what it does not support. Verified 2 September 2026.
The 12 legal signals, side by side
Recorded rather than graded. These are the questions a practitioner has to answer before a tool touches a client matter, and the answers are taken from public material only.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
The published terms of service, in force from 12 November 2025, state that the provider is entitled to use the history of queries a user submits during interaction with the AI in order to develop the AI algorithms through which it supplies the services. No opt out is located. The qualifier the vendor attaches is recorded here: the clause states that query histories will not contain third party personal data disclosed by the user, which is supported elsewhere in the same agreement by a bar on submitting third party personal data without a signed data processing agreement.
This record notes that the English marketing estate separately states that client files, cases and analyses are never used to train any AI model by the vendor or any third party. The grant is in the agreement and the denial is in marketing; the index records both and concludes neither.
Checked the home page, product page, pricing page and its FAQ, the terms and conditions in full, the privacy policy in full and the trust center index on 2 September 2026. Nothing addresses whether customer content is used to train or improve models, in either direction. The terms are silent: clause 7 covers data protection by reference to the privacy policy and deletion after contract end, and clause 8 reserves rights in the software, content, models and generated data to the provider without touching training.
The privacy policy is expressly website-scope, stating that use of the application at app.omnilex.ai is additionally governed by the data protection terms of the license, in particular the data processing agreement documented in the trust center. That agreement exists as a license agreement including a DPA template listed among the trust center documents, behind an access flow, and it was not requested. Two adjacent statements point toward a protective posture without reaching the question: models are stated to be processed on Swiss infrastructure so that attorney-client privilege stays intact, and the license is described as drafted to respect professional secrecy and auxiliary-person liability.
Neither names training. Under the naming test the honest record is that the question is unaddressed on every public surface.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
Retention of query history is acknowledged in several places in the terms without any period being stated. The provider undertakes to let a user export content stored under the services, named as query history, before an account is removed; an account inactive for over a year with no subscription may be deleted; and account deletion on request is completed within 24 hours, extending to 14 days where there are technical obstacles.
No retention window is published for prompts, outputs or uploaded documents while an account is live, and no zero retention setting is described.
Retention is acknowledged in the operative terms and no period is attached to it. Clause 7 states the quoted position, which fixes an event rather than a duration: deletion follows the end of the contract, with no stated window for how long after, and with an open carve-out for statutory retention duties. Nothing states how long prompts, chats, generated drafts or uploaded documents persist during the term, and no retention setting is offered to the customer.
The trust center lists a data erasure item and a data lifecycle management control among the defaults, and the home page names data lifecycle management as one of the controls legal teams expect, but neither carries a figure on any surface that could be read. The one specific limit published anywhere concerns size rather than time, with the pricing page capping documents at 1,000 pages and 500MB. The data processing agreement that would normally carry retention detail sits behind the trust center access flow and was not requested.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
No located material addresses ethical walls or matter level segregation. The question arrives differently here than for a document management integrated product, because this product has no such integration: documents are uploaded into the vendor's own store rather than retrieved from the firm's system, so there is no existing access model for it to inherit. Team plans are described in the terms as pooling a shared allowance across seats under an administrator, with no per matter or per user access boundary described. Checked 15 September 2026 across both estates, the terms of service and the privacy policy.
Checked the home page, product page, pricing page, terms, privacy policy and trust center index on 2 September 2026. Nothing describes segregation between customers, between users inside a firm, or between matters. The product is visibly organized around matters, with the interface showing named projects such as a share purchase, a due diligence, a litigation matter and an estate, and the pricing page adds team and project collaboration at the Team tier and know-how integration, so the structure for separation plainly exists.
What is absent is any statement that it functions as a boundary: nothing says whether one user in a firm can reach another's project, whether ethical walls can be configured, or how tenants are isolated from one another. Access controls are published at the account layer instead, with SAML single sign-on, audit logs and IP allow-listing. The gap sits awkwardly against the vendor's own professional-secrecy framing, since conflicts walls are part of what Swiss professional secrecy requires of a firm.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
No commitment to notify a customer of a subpoena, law enforcement demand or other third party request for its data was located, and no discretion over notice is reserved either. The terms of service were reviewed in full and contain no clause on the subject. No transparency report is published. Checked 15 September 2026.
Checked the terms and conditions in full, the privacy policy in full, the home page, product page, pricing page and trust center index on 2 September 2026. Nothing addresses disclosure to authorities or in response to legal process, and nothing addresses notice to the customer if a demand arrives. The terms have no confidentiality article at all and no required-disclosure clause. The privacy policy lists legal obligations as a processing basis but is expressly scoped to the website rather than the application, and its recipients are website service providers.
The question is therefore not reached rather than answered adversely. Two things are recorded so a later grader can weigh the gap: the license agreement including a DPA template is listed among the trust center documents behind an access flow and was not requested, and it is the document where a compelled-disclosure clause would sit; and the vendor's professional-secrecy framing makes the silence more conspicuous, since a demand for a lawyer's file is precisely the scenario auxiliary-person liability contemplates.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Sources are identified with unusual specificity for this corpus. Poland is described as statutes, common and administrative court rulings, tax interpretations, the KRS company registry, the Eureka jurisprudence database, Constitutional Tribunal and Supreme Court rulings; the Czech Republic as the Civil Code by act number plus Constitutional and Supreme Court rulings; Germany as BGB, HGB and BGH, BAG and BVerwG jurisprudence; Austria as ABGB, UGB, AVG and Oberster Gerichtshof rulings.
A daily update cadence is stated. No license or rights basis is stated for any of it, although the terms describe results as depending on the availability of data gathered from publicly accessible websites publishing rulings, which points at a public source basis without asserting one.
The corpus is identified by category and counted to the unit, which is more precise than any other record in this pull: 797,321 court decisions, 222,525 commentaries, 22,181 laws and articles, 13,162 official documents and 2,352 pieces of literature, totaling 1,057,541 sources, described as federal and cantonal decisions, statutes, official documents and literature refreshed continuously. Cadence is therefore described and the composition is transparent.
One category carries its own provenance disclosure and deserves credit: the Omnilex Commentary is the vendor's own, written article by article, and it is labeled as AI-drafted and cited back to statute and case law rather than presented as published doctrine. What is not stated anywhere is the rights basis. No publisher, database or data supplier is named for any category, and no license, agreement or permission is described.
That gap matters most for the 2,352 pieces of literature and for whatever underlies the commentary category, since Swiss legal literature is commercially published and is not, unlike federal and cantonal case law, free to redistribute by default.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
The product verifies that a cited ruling exists, by checking case signatures and documents against its database, and compares lines of jurisprudence to identify a dominant judicial position. Neither is a treatment check. No located material states whether authority the product returns is tested for subsequent history, and no citator is named. The distinction matters on this record specifically, because existence verification is marketed prominently and a reader could take it for currency verification.
Checked the home page, product page, pricing page, terms and trust center index on 2 September 2026. No public material addresses whether a decision returned by the product is still good law. Nothing describes treatment flags, subsequent history, or whether a judgment has been confirmed, distinguished or departed from. The nearest claims concern the freshness of the collection rather than the standing of any individual authority: sources are described as refreshed continuously, and the product page offers to keep a user informed about the latest changes in legislation and jurisprudence at federal, state and cantonal levels.
Those speak to what has been added, not to what has been overtaken. The question bites here because the product retrieves case law and its own commentary as the basis for cited answers, and because Swiss practice depends heavily on whether the Federal Supreme Court has since moved.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
The vendor documents an explicit abstention behavior: the system checks whether a signature or document exists in the database and, where confirmation is absent, flags that rather than presenting the answer as certain. The vendor's own phrasing is that it does not pretend to be sure. This is a described behavior with a stated check rather than a disclaimer, which is why it records as documented. It is not demonstrable, because no published evaluation or worked example of the refusal path was located.
Checked the home page, product page, pricing page, terms and trust center index on 2 September 2026. Nothing describes what the platform does when it cannot ground an answer, and no confidence, relevance or grounding indicator is described as shown to the user. The published answer to the same underlying risk is verification by the reader rather than abstention by the system: every claim is stated to be cited, the commentary carries the line that every statement is citable, and clause 6 of the terms places responsibility for reviewing and adapting all content on the user.
The user can also restrict which source types are consulted, which shapes retrieval rather than describing behavior when retrieval comes back thin.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
No court order, opinion or disciplinary record addressing fabricated or hallucinated legal citations in output from this product was located as of 15 September 2026. The tracker searched was the AI Hallucination Cases database maintained by Damien Charlotin, alongside a general search of the public record. This is a statement about the public record on that one subject as of that date. It is not a finding about the product, and this signal is not a litigation history.
Searched the AI Hallucination Cases database maintained by Damien Charlotin, and reporting drawing on it, on 2 September 2026 on the product and corporate name Omnilex and Omnilex AG. No court order, opinion or disciplinary record naming the product was located. This is a statement about the public record rather than a finding about the product. One structural caveat bounds it: the database is heavily weighted to United States filings, its European coverage is thinner, and Switzerland does not feature among the jurisdictions commonly reported in it, so a Swiss product is less likely to surface even where an incident occurred.
Swiss judgments are also published in three languages, which further limits how reliably an English-language search reaches them.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
The terms of service engage with professional responsibility in general terms without naming any guidance. They state that the services are intended to support persons professionally providing legal assistance who have the knowledge and experience to use the results properly, that the services in no way replace legal assistance provided by professionals, and that a user taking any legal step should consult someone professionally qualified and holding the appropriate professional title.
No bar or ethics opinion is named, in Poland or in any of the other three jurisdictions the product covers, and no material is published that a firm could hand to its own risk committee.
Professional obligations are referred to directly but no authority or instrument is named. Omnilex states that the platform was built for professional secrecy from day one, that models are processed on Swiss infrastructure so attorney-client privilege stays intact, and that its license agreement and data processing agreement are drafted to respect professional secrecy and auxiliary-person liability. Naming auxiliary-person liability is a substantive reference to the doctrine governing a Swiss lawyer's responsibility for third parties they engage, and it is more than most records in this corpus offer.
What is missing is the source. No instrument is cited, with neither the Federal Act on the Free Movement of Lawyers nor the criminal provision on professional secrecy named, and no guidance from the Swiss Bar Association or any cantonal bar is referenced. Nothing addresses what a lawyer must do to satisfy those duties when using the product, as opposed to what the vendor has done to make it possible.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
The vendor markets compressed time to the buyer, offering tax interpretations and dominant jurisprudence lines in minutes and a complete argument built in minutes, and the product is sold to law firms and solo practitioners who bill clients for that work. No located material addresses billing treatment, fee disclosure, or what a firm should tell a client about AI assisted work. The Lex consumption table gives a firm a precise record of what each action cost it in units, but that is an internal metering record priced to the firm and not a record of AI assisted work framed for a client bill.
Checked the home page, product page, pricing page and FAQ, terms and trust center index on 2 September 2026. Nothing addresses how AI-assisted work is recorded, billed or disclosed to a client, and no per matter record of AI-assisted work was located. Unusually for this corpus, no quantified savings claim appears on any first-party surface either: the vendor's language is about being faster, more precise and dependable, and about winning the matter rather than the admin, without publishing a time or cost figure, so the weakest value is not made out.
A third-party report attributes a 93 percent research time saving to user reports; that is an excluded source and is not recorded. The raw material for a per matter record does exist in the product, which organizes work into named mandates and carries audit logs, but nothing describes either as reportable against a bill.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
A current model provider list is published in the privacy policy, naming OpenAI, OctoML, Gladia, Anthropic and RunPod as the providers of the engines powering the AI, and stating that each processes under a data processing agreement incorporating European Commission approved standard contractual clauses. That is the artifact a firm needs when a client's outside counsel guidelines ask which model providers see its content, and it is reachable without an agreement in place.
It falls short of a disclosure pack because no client facing consent or notification material is published alongside it, and the list names providers rather than models.
A subprocessor list is published and reachable without an agreement: the trust center legal section identifies Amazon Web Services and Microsoft Azure, and the same portal carries a data processing agreement item, a cyber insurance item and a license agreement including a DPA template. The pricing FAQ confirms that a data processing agreement is part of every license. That is enough for a firm to begin answering a client, which is why the bottom value is false here.
It stops short of the top value on the limb this signal treats as decisive. The two named subprocessors are infrastructure, and infrastructure alone does not answer whose model saw the content: five frontier models are named on the home page as selectable, and not one of their providers appears in the list or anywhere else as a contracting entity. The forwardable material is also gated rather than published, with the license and DPA sitting behind the trust center access flow, so what a firm can hand to a client without asking is the subprocessor naming and the certification rather than the documents themselves.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
No located material addresses court disclosure obligations, judicial standing orders or verification certification, and no export covering model used, sources retrieved and human verification is described. A full audit trail is claimed for the judiciary segment, but it is described as organizing case material for a judge rather than as a record of AI assistance a filer could disclose. Checked 15 September 2026 across both estates, the terms of service and the privacy policy.
Checked the home page, product page, pricing page, terms and trust center index on 2 September 2026. Nothing addresses disclosure of AI use to a court or authority, and no exportable per document record of model used, sources retrieved and human verification is described. The product does leave a usable trail for the lawyer's own purposes, since every claim is cited and ready for the file, the commentary is labeled as AI-drafted, and audit logging is published as a standard control, but none of these is presented as a record of model use and none captures which of the five selectable models produced a given passage.
That last point is the sharpest version of the gap on this record: a platform that lets a user choose between five frontier models, and that labels its own commentary as machine-written, is closer than most to being able to emit exactly this record, and does not say that it does.
The questions both sides leave open
Derived from the records above rather than written, so it cannot favor either vendor. Take these into both conversations and ask each side the same question.
- Ethical Walls and Matter Segregation
- Third Party Request and Subpoena Notice
- Good Law Verification
- Court Disclosure Support
Which one fits
Choose Gaius-Lex if
- You practice Polish, Czech, German or Austrian law. Gaius-Lex retrieves from a daily updated database naming the primary sources for each, from Polish tax interpretations and the KRS registry to BGH and Oberster Gerichtshof rulings, with footnoted links and case signatures checked for existence.
- You need the model to run on your own servers. Gaius-Lex offers SaaS, hybrid and on premise deployment as published buyer choices rather than enterprise conversations.
- You want to price it and try it alone. Gaius-Lex publishes Gaius Standard at 159 euro and Gaius Pro from 2,150 euro a month net, a seven day trial without a card, and a unit, the Lex, with a published consumption table.
Choose Omnilex if
- You practice Swiss law. Omnilex counts its corpus openly at 797,321 court decisions, 22,181 laws and articles and more than 222,000 commentary entries of its own, labeled as AI drafted and cited back to statute and case law.
- You want to choose the model and keep it in Switzerland. Omnilex names GPT-5.6, Claude Opus 5, Mistral Large 3, Kimi K2.6 and Grok 4.6 as selectable, processed on Swiss infrastructure hosted on Microsoft Azure in Zurich.
- Your security review expects a certificate. Omnilex holds ISO 27001:2022 across the organization, and its trust center lists a dated penetration test report, cyber insurance and a license agreement with a data processing agreement template.
In summary
Gaius-Lex
Gaius-Lex, operated by FLATHUB sp. z o.o. of Krakow and backed by Digital Ocean Ventures, is an AI research and drafting agent for lawyers and tax advisers covering Polish, Czech, German and Austrian law, retrieving from a daily updated database of statutes, case law and tax interpretations with footnoted citations. The AI Legal Index grades it in the top two bands on nine of fifteen capability axes, with A grades on AI centrality and pricing. It offers SaaS, hybrid or on premise deployment, names OpenAI, Anthropic and three other providers in its privacy policy, and prices from 159 euro a month. As of 15 September 2026 the index located no named customer, security certification or integration.
Omnilex
Omnilex, from Omnilex AG of Zurich, is an AI workspace for Swiss law covering more than a million federal and cantonal sources, its own AI drafted commentary, and no code agents built from a firm's playbooks, with the user choosing among named frontier models processed on Swiss infrastructure. The AI Legal Index grades it in the top two bands on thirteen of fifteen capability axes, with an A on pricing. It holds ISO 27001:2022, publishes per seat prices from CHF 1,600 a year, and names the Swiss Confederation, Bayer and several law firms among customers. As of 2 September 2026 the index located no training position or named integration.
Questions buyers ask
Gaius-Lex vs Omnilex: which is better for European legal research?
It depends first on the jurisdiction: Gaius-Lex covers Polish, Czech, German and Austrian law and Omnilex covers Swiss law. On the AI Legal Index grid Omnilex sits in the top two bands on thirteen of fifteen capability axes and Gaius-Lex on nine of fifteen, identical on eight. Omnilex publishes more on security, customers and professional secrecy; Gaius-Lex offers on premise deployment.
Does Gaius-Lex train on client queries?
Its published terms, in force from 12 November 2025, reserve a right to use query history to develop its AI algorithms, stating that histories will not contain third party personal data. Its English site states that client files, cases and analyses are never used to train any model. The terms are the binding text. Omnilex publishes no position on training. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 27, 2026. No vendor pays for placement.
Which models do Gaius-Lex and Omnilex use?
Omnilex names five selectable models to the version, GPT-5.6, Claude Opus 5, Mistral Large 3, Kimi K2.6 and Grok 4.6, processed on Swiss infrastructure. Gaius-Lex's privacy policy names five providers, OpenAI, OctoML, Gladia, Anthropic and RunPod, under standard contractual clauses, without naming models. Neither commits to notice when the list changes. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 27, 2026. No vendor pays for placement.
How much do Gaius-Lex and Omnilex cost?
Gaius-Lex publishes Gaius Standard at 159 euro a month net and Gaius Pro from 2,150 euro a month net, metered in a unit called a Lex, with a seven day trial. Omnilex publishes CHF 2,000 per user a year for one user and CHF 1,600 per user a year for teams, with onboarding and training included and a fourteen day trial. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 27, 2026. No vendor pays for placement.
What do Gaius-Lex and Omnilex both leave unpublished?
A good law check and a named integration. Neither checks whether a returned decision has since been overruled, and neither names a document management system it connects to. Neither describes segregation between matters inside a firm, and neither publishes an accuracy measure for its answers. Graded by AI Legal Index against 15 capability axes and 12 legal signals, including privilege handling and citation accuracy, from each vendor's own published materials, verified September 27, 2026. No vendor pays for placement.
Three readings to weigh. Gaius-Lex's terms permit using query history to develop its AI, while its English site says client files never train any model; its site also says documents stay in the European Economic Area while its privacy policy names US providers among recipients. Omnilex publishes nothing on training either way, and its data processing agreement sits behind a request. The two products cover different national laws. Gaius-Lex was verified on 15 September 2026 and Omnilex on 2 September 2026. Neither vendor reviewed this page.
Neither vendor paid for inclusion, placement or a grade, and neither reviewed this page before it published. Everything above comes from public material on the dates shown. How the index grades.