M
Minesoft PatBase

Minesoft PatBase is a global patent search and analytics platform used by more than 90,000 IP, legal and R&D professionals for prior art searching, freedom-to-operate work, portfolio management, competitor monitoring and landscape analysis. It draws on more than 180 million patent publications from over 100 issuing authorities, organised at both family and publication level, with full-text coverage of the major jurisdictions, daily quality checking of the underlying data and a live feed of new publications.

Searching runs from natural language through to complex Boolean and command-line syntax, and a spring 2026 release added fourteen standard-setting organisations to the platform for standard-essential patent work, among them IEEE, ISO, ITU-T, IETF and ANSI, searchable by a dedicated syntax. Several capabilities are AI-driven and most arrived across 2025 and 2026. Document Chat lets a user ask targeted questions of an individual patent and receive generated answers, included in all plans at no extra cost.

AI Explain highlights the precise sections and wording of a document that overlap with the user's query, so a searcher can see why a result appeared. Smart Images applies image recognition to patent drawings, identifying reference numerals and mapping them to the corresponding passages of the description in an automatically generated parts list. AI alerts can be bulk-trained from a user's own folders and search history to sharpen a classifier against their actual work.

Natural-language queries are expanded by AI and results carry relevance scores, and the Smart Claims viewer distinguishes independent from dependent claims. Machine translation is embedded throughout, developed with RWS, the translation company that co-develops PatBase with Minesoft and led its move to cloud hosting. Minesoft Ltd is a privately held British company founded in 1996 and based in Richmond-upon-Thames, London, whose other products include Minesoft Origin, PatDocs and ChemX, alongside a patent data API.

Vendor siteRichmond-upon-Thames, London, United KingdomFounded 1996
Last verifiedSeptember 13, 2026
Compare with other vendors

Capability grades

All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.

BB on AI CentralityThe models are the engine of a core capability, layered on a product that would still function without them as a document or workflow system.

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 models are the engine of several named capabilities layered on a platform that functions fully without them, which is the B band. The AI additions are real and recent, arriving mostly across 2025 and 2026: Document Chat for generated answers about an individual patent, AI Explain for showing where a query overlaps a document, Smart Images for reading reference numerals out of patent drawings, AI-trainable alert classifiers, natural-language query expansion with relevance scores, the Smart Claims viewer distinguishing independent from dependent claims, and machine translation embedded across search and review.

Underneath sits a platform whose independent existence is not in doubt and is in fact the vendor's oldest claim on its buyers. PatBase has been in market since 2003 on a company founded in 1996, and its core is a curated corpus of more than 180 million publications from over 100 authorities with family-level organisation, daily data quality checking, and Boolean, proximity, wildcard and command-line searching that the vendor still markets to professional searchers as a primary mode.

Strip out every model and a searcher retains the corpus, the family structure, the classification tools and the syntax. What the AI changes is who else can use it and how quickly, which is the B band exactly. Verified 13 September 2026.

Source: Vendor Published
BB on Citation Accuracy and Hallucination DisclosureGrounding is real and documented, with linked primary sources and a described retrieval method, short of published accuracy figures an outsider can test.

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 real and documented at an unusual level of granularity, with no measured accuracy published, which is the B band. The grounding mechanism is a named feature rather than an architectural assertion. AI Explain shows the user precisely where and how their input overlaps a document's content, highlighting the specific sections and wording that matched, and the vendor states the purpose plainly: so a searcher can immediately judge how relevant a result is and understand why it appeared.

That is passage-level attribution on a retrieval result, which is more than most records in this lane offer. Relevance scores accompany results, giving a ranked signal a user can act on, and Document Chat operates against a single patent so its answers have a bounded and visible source document. What is absent is measurement of the vendor's own accuracy. No precision or recall figure is published for the AI search, no test set is described, no evaluation of Document Chat is offered, and no failure mode is named.

The agreement runs the other way and is worth recording because it is contractual rather than promotional: clause 11 disclaims any warranty that the Information is accurate or complete, and states that the licensor does not guarantee the accuracy of the Product or the Information or accept liability for the consequences of their use. Verified 13 September 2026.

Source: Vendor Published
BB on Autonomy and Oversight ModelA written commitment that the models work alongside a supervising lawyer, with real review surfaces, short of the full control structure: commonly the threshold at which the system stops or what happens after it is wrong.

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.

Real user control over how the models behave, with review surfaces, short of the full control structure, which is the B band. Two published capabilities go beyond assurance to actual control. AI Explain lets a user interrogate why a result surfaced, exposing the overlap between query and document rather than presenting a ranking as a verdict. And AI alerts can be bulk-trained by the user from their own folders and search history, so the classifier that decides what reaches them is tuned by their own prior judgements rather than by the vendor alone.

That is user authority over model behaviour, which is rarer in this corpus than statements about human oversight. Relevance scores give a graded signal instead of a binary inclusion, letting a searcher decide where to stop. What the A band asks for is not published. No threshold is stated at which any capability acts without review, no mode distinction separates AI-assisted from conventional operation despite both being available, nothing describes what happens when a Document Chat answer is wrong, and no route is published for correcting or reporting an output.

Nor is there any general statement that the tools support rather than replace professional judgement, which several comparable vendors do make. The controls here are practical rather than declared, and the note records that as the shape of the record. Verified 13 September 2026.

Source: Vendor Published
CC on Operational and Outcome EvidenceCustomer logos and unattributed testimonials stand in for evidence, or results are quoted with no basis stated.

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.

Scale and longevity are published without a single named customer or measured outcome, which is the C band. What exists is substantial as scale evidence. The vendor states more than 90,000 innovators and legal professionals use the platform, that it has been helping patent professionals since 1996, and that PatBase itself launched in 2003, so the operating history is long and stated. Corpus scale is quantified at more than 180 million publications across over 100 authorities.

Support reach is described concretely at 15 or more languages across all time zones. A public product status page is maintained, which is a small but real piece of operational transparency few vendors in this lane offer. None of that is deployment evidence in the sense this axis tests. No customer is named anywhere on the surfaces read, no case study or customer story section appears in the navigation, no testimonial is attributed to a named individual or organisation, and no figure is published for time saved, recall improved or cost reduced by any of the AI capabilities.

The feature announcements describe benefits in the abstract, saying that a capability is faster or produces less noise, without measurement. The one third-party attested marker located, a corporate identity and registration number, evidences the company rather than the product. Verified 13 September 2026.

Source: Vendor Published
BB on Privilege and Confidentiality PostureSubstantive published commitments on confidentiality and training use, short of the full picture: commonly silence on segregation between users or matters, or on what the underlying model provider may retain.

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.

Substantive published commitments, several of them contractual, defeated on the limbs the A band names, which is the B band. The commitments are unusually direct for a vendor of this size and they come from two independent places. On the product surfaces: customer data is never used to train models, searches and results are stated to belong to the customer, and the platform is described as a private cloud with data stored securely and privately in compliance with data protection law.

In the published agreement: clause 19 imposes a mutual confidentiality obligation surviving five years beyond termination, restricts use of the other party's confidential information to performance of the agreement, and requires notice before any legally compelled disclosure so far as the law permits, which is graded on its own signal. Clause 20 commits both parties to the UK Data Protection Act 2018 and the GDPR. Clause 4 gives the customer a right to retain an archive copy of retrieved information for twelve months after termination.

Three limbs fail. Privilege and work product are not addressed by name, which under R33 forecloses A on its own. No model provider is identified for the generative features. And no retention or deletion position is published for the customer's own searches, queries or Document Chat exchanges, which is the material a competitor would most want. Verified 13 September 2026.

Source: Vendor Published
CC on UPL and Professional Responsibility PostureA boilerplate disclaimer sits in the terms while the marketing describes the product in advice terms, or the intended audience is left ambiguous.

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.

A contractual disclaimer sits in the terms while the marketing describes the product in decision terms and the audience reaches beyond lawyers, which is the C band. The disclaimer is real and is contractual rather than a website footer: clause 11 provides the Product and the Information are supplied as is, disclaims any warranty that the Information is accurate or complete, and states the licensor does not guarantee accuracy or accept liability for the consequences of use.

Clause 10 adds that features dependent on public patent office services are outside the licensor's control and no responsibility is taken for the successful retrieval of Information, which is a pointed disclaimer on a prior art tool. Against that the product is marketed for freedom-to-operate analysis, patent due diligence and invalidity screening, all of which produce legal conclusions, and the vendor addresses legal teams, IP experts, professional searchers and R&D teams together, so non-lawyers are expressly among the intended users.

Nothing published bridges the two. No statement addresses where the tool's output stops and professional judgement begins, no supervision or competence dimension is offered, no rule of professional conduct or bar authority is named in any jurisdiction, and nothing addresses an R&D user acting on a Document Chat answer about claim scope without counsel. Verified 13 September 2026.

Source: Vendor Published
CC on AI Governance and Bias DisclosureResponsible AI principles are published without a mechanism, a testing regime, or anything a buyer could audit.

AI Governance and Bias Disclosure

Published governance over model behaviour: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.

Individual AI commitments are published without any governing framework behind them, which is the C band. What exists is a set of true and useful statements rather than a programme. Customer data is never used to train models. Processing takes place in a private cloud on UK-based servers. Data centres are said to meet ISO 27001. AI Explain is presented explicitly as a transparency feature, and the autumn 2025 release framed a group of changes around enhancing transparency, so the vendor does think about explainability as a design goal and ships features that serve it.

Beyond that the governance layer is absent. No AI policy or responsible-AI page was located in the site navigation, and the dedicated AI section of the estate was not read and is named under R25 as the surface most likely to change this grade. No framework is adopted or cited, no individual, committee or function is named as accountable for model behaviour, no pre-release testing regime is described and no evaluation result is published.

Bias is not addressed in any form, which is worth naming on this product specifically because the alert classifier is trained on an individual user's own prior selections, a design that by construction reinforces what that user has already chosen to look at, and nothing published addresses the narrowing that invites. Verified 13 September 2026.

Source: Vendor Published
BB on AI Safety and Data StewardshipSubstantive published policy covering most of the ground, short of the full set: commonly no named subprocessor list or no stated incident practice.

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.

Substantive published policy across most limbs, short of the full set, which is the B band. The data position is stated plainly and repeatedly across the product estate: customer data is never used to train models, searches and results belong to the customer, information is stored securely and privately within a private cloud, and servers are located in the United Kingdom. Security certification is claimed against ISO 27001 for the data centres, and the agreement adds obligations on both sides, clause 6 making the customer responsible for the security of issued User IDs and requiring immediate notification of any unauthorised usage, with a reciprocal audit right for the licensor on seven days' notice.

Clause 19 carries a five-year confidentiality tail and clause 20 commits to the UK Data Protection Act 2018 and the GDPR. A public product status page is maintained, giving customers visibility of availability without asking. Three limbs fail. No subprocessor list is published, and the one third party identified anywhere, RWS as machine translation partner, is named in a development context rather than as a processor of customer data.

No retention period or deletion commitment is published for searches, queries or Document Chat exchanges. And no incident or breach notification commitment to customers was located, the agreement placing the notification duty on the customer and not reciprocating. Verified 13 September 2026.

Source: Vendor Published
BB on AI Liability and RecourseA real published position on liability, short of the full picture: commonly a stated indemnity without scope or caps.

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.

A published agreement addressing every commercial limb the axis tests, short of anything AI-specific, which is the B band and the strongest liability position in this lane. The agreement is published in full as a downloadable document, ungated, and it is short enough to read, which by itself distinguishes it from the three comparators in this lane whose terms could not be recovered at all. An express warranty exists and is not swallowed by the disclaimer that follows it: clause 11 warrants that the Product will operate materially in accordance with its description or technical specification, before disclaiming all other warranties.

Liability is capped at the licence fee paid in the preceding twelve months, with the usual exclusions for loss of profits, revenue or business and for indirect, consequential, special, punitive or exemplary loss. And a vendor indemnity is given rather than withheld: clause 13 requires the licensor to indemnify the customer against all claims, damages and expenses arising from a claim that the Product infringes a third party's intellectual property rights, subject only to notice, control of defence and no-admissions conditions, with a reciprocal customer indemnity for use of the Product.

Termination runs on material breach with fourteen days to remedy, and English law and courts govern. What is absent is any AI provision at all. Nothing addresses the accuracy of a generated answer, nothing allocates risk for a decision taken on one, and no insurance position is published. Verified 13 September 2026.

Source: Vendor Published
CC on Practice Systems Integration DepthIntegrations are listed as logos or marked as coming, with no documentation an implementer could use.

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.

Integration surfaces are named as products without documentation or a named external system, which is the C band. What is established is that routes out of the platform exist and are sold as such. A Patent Data API is published as a product line in its own right, so programmatic access is offered rather than merely implied. The estate lists an MCP entry among its AI applications, which if shipped would let a customer's own assistant call the platform, and under ground rules section 3 that is extensibility rather than model choice; it is recorded here and expressly not credited, because the page was not read and the estate simultaneously advertises an early-access programme for an agentic tool, so whether MCP is generally available or part of that programme is unresolved.

Within the suite, PatBase sits alongside Minesoft Origin, PatDocs and ChemX, and analytics output including charts, diagrams, heat maps and clusters can be exported and shared. What the higher bands require was not established. No connector catalogue is published, not one external system is named as supported, no API documentation was reached, and nothing describes how results or portfolio data would reach a docketing or IP management system. Under R25 the API and MCP pages are named as what would move this row. Verified 13 September 2026.

Source: Vendor Published
BB on Deployment Model and Data ResidencyDeployment model is stated clearly with partial residency detail, or residency is offered without the processing location being addressed, or the tenancy model is stated on its own with no residency detail published.

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.

Both co-equal limbs are addressed at least in outline, which under R38 places this comfortably in the B band and above most of its lane. Region is stated first-party and repeated across the estate: the servers are based in the United Kingdom, offered as a reason a customer's data is secure. For a European buyer with residency obligations that is a direct answer, and for a buyer needing United States or Asian residency it is an equally clear one in the other direction, which is the point of the axis.

Tenancy is addressed in substance if not in architectural terms: the platform is described as a private cloud, with a customer's searches and results characterised as their own and held within it, and the 2020 platform rebuild is described as a cloud implementation led by RWS aimed at resilience and security. Data centres are stated to meet ISO 27001. What is not published keeps it off A. No deployment options are offered or refused, so nothing states whether a single-tenant, private or on-premises arrangement is available to an enterprise that needs one, and no alternative region is offered for customers who cannot use United Kingdom hosting.

The private cloud description states an outcome rather than an architecture, and nothing sets out how separation between customers is achieved or assured. No data centre operator is named. Verified 13 September 2026.

Source: Vendor Published
CC on Security Certifications and Trust CenterBadges appear on the site with no scope, no date, and no report available.

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.

A standard is named and the claim is made in terms that leave its subject unclear, with no scope, date or evidence, which is the C band. The sentence is the whole of the disclosure and it repays close reading: the vendor's commitment to security is said to be backed by ISO 27001 certification, ensuring that its data centres adhere to international information security standards. Read strictly, the certification is attached to the data centres rather than to Minesoft's own information security management system, which raises the question ground rules section 3 exists for, since a hosting provider's certification is infrastructure and does not attest to the vendor's own controls.

Nothing published resolves which is meant. Beyond that ambiguity, none of the supporting material a buyer would use is present: no certificate number, no certification body, no issue or expiry date, no statement of the scope covered, and no route to obtain the certificate or any report. No trust centre or security documentation page exists in the site navigation. What sits alongside is real but is practice rather than attestation, being the private cloud description, the United Kingdom server location, and the User ID security and audit provisions of clause 6 of the agreement.

The distance to a better grade is small and entirely within the vendor's control, which the note records because it is the honest reading. Verified 13 September 2026.

Source: Vendor Published
CC on Model Supply Chain DisclosureThe vendor refers to advanced or proprietary models without identifying what sits underneath.

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.

The vendor describes AI capabilities by function without identifying what performs them, which is the C band. Seven distinct AI features are named and described: Document Chat, AI Explain, Smart Images, the trainable alert classifier, natural-language query expansion with relevance scoring, the Smart Claims viewer and machine translation. For each, what it does is clear and how it is built is not. No model is named, no version is given, no architecture is described beyond references to AI-based image recognition and AI-enabled claim recognition, and no provider is identified for the generative component, which matters most on Document Chat because generated answers about a patent are where a third-party foundation model would ordinarily sit.

Nothing states where inference runs relative to the United Kingdom servers the vendor otherwise emphasises, what any provider retains of a question put to Document Chat, or whether a customer would be told if an underlying model changed. Clause 10 of the agreement runs the other way, reserving the right to change or modify the Product and its features without notice. One third party is named and is credited here as far as it goes: RWS, the translation company that co-develops the platform, is identified as the machine translation partner and is said to be continuously improving translation quality.

That is a named supplier performing a named function, which is real disclosure, and it covers translation only. Verified 13 September 2026.

Source: Vendor Published
CC on Commercial TransparencyPricing is gated behind a demo request while tier names and feature splits are published, so the shape is visible and the number is not.

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.

The unit and structure of charging are published in the agreement while no figure appears anywhere, which is the C band and lifts this off the floor. The published terms set out the commercial mechanics with precision, and a buyer can establish the shape of the deal before speaking to anyone. Licensing is per User ID, with additional User IDs added on written request, co-terminous with the existing term and charged pro rata by the month in which they are added.

The term is twelve months unless the Product Terms say otherwise, renewing automatically for further one-year periods unless either party gives thirty days' written notice. Invoicing is annual in advance. Late payment attracts interest at eight per cent above the Bank of England base rate together with collection costs, and the licensor may suspend access without notice while payment is outstanding. Clause 17 provides that Product Terms prevail over the general Terms in a conflict, so a buyer knows where the commercial specifics will live.

What is absent is any number. No price, band, minimum or indicative range appears on the estate, and the only route published is a demonstration request. Under R17 the test is mechanical: this evidence lifts the axis above D, so a VendorPricing row is owed, written with entryPriceUsd null under R10. Verified 13 September 2026.

Source: Vendor Published
BB on Firm and Practice CoverageSegment and practice coverage is described with substance, short of the boundaries: what is supported is clear, what is not is left open.

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.

Coverage is described with real substance across data, jurisdictions and workflows, with the boundaries left open, which is the B band. Corpus coverage is quantified at more than 180 million publications from over 100 issuing authorities, with full-text coverage of the major jurisdictions, family-level and publication-level organisation and a live feed of new documents subject to daily quality checking. Language coverage is addressed through machine translation embedded in search and review, often available the same day as a new document, which for cross-border prior art is coverage in the operative sense.

One area is covered in unusual and recent detail: standard-essential patent work, where a spring 2026 release added fourteen standard-setting organisations by name, being IEEE, ISO, CCSA, OMA, IMT, OneM2M, TIA, ATIS, ITU-T, ITU-R, IEC, IETF, ANSI and TTA, with dedicated search syntax and new source and company columns. Workflow coverage is published as named use cases across search, monitoring, collaboration, due diligence, freedom to operate, competitor intelligence and technology landscaping, and buyer coverage names IP experts, legal teams, R&D teams and professional searchers, supported in fifteen or more languages.

What is left open is every limit. No jurisdiction is identified as thinly covered among the hundred, no technology area is named as better or worse served, and nothing states coverage dates or backfile depth. Verified 13 September 2026.

Source: Vendor Published
Sources on file

4 public documents

The public pages on file for Minesoft PatBase, with the recorded signals each one supports and the date it was last read. Open any of them and check the reading against the record.

Pricing

No published figure

  • Minesoft does not publish what PatBase costs, but unusually it does publish how the charging works, in terms you can download before you talk to anyone.
  • You licence by named user. Each person gets their own User ID, and you cannot share one — the agreement is explicit that a User ID may not be used by anyone who is not your employee. If you add people mid-year, they join on the same end date as everyone else and you pay a pro-rated amount from the month they start.
  • The standard term is twelve months, invoiced a year in advance. It renews automatically for further twelve-month periods unless you or Minesoft give thirty days' written notice before the end of the current term. Miss that window and you are committed to another year.
  • Late payment is handled firmly: interest runs at 8% above the Bank of England base rate after thirty days, you pay collection costs, and Minesoft can suspend your access without notice while an invoice is outstanding.
  • One thing worth knowing if you are budgeting for the end of a subscription rather than the start: you may keep an archive copy of what you retrieved for twelve months after the agreement finishes, but you may not hold it in a searchable database or share it outside your organisation without written permission and, if required, an extra fee.
  • The actual numbers live in a separate Product Terms document that is not published, and that document takes precedence over the general terms if the two conflict. So expect the price, and any product-specific commercial conditions, to arrive only through a sales conversation.

No figure is published at any level. The unit of charge and the billing structure are published in the UK and International Annual Subscription Terms, which is what carries this above the floor. Licensing is per User ID, non-exclusive and non-transferable, with User IDs not permitted to be used by personnel not employed by the customer (clause 5(e)). Additional User IDs may be added on written request, take the same initial or renewal term as the existing agreement, and are charged pro rata based on the month of addition (clause 8).

The term is twelve months unless the applicable Product Terms specify otherwise, renewing automatically for successive one-year periods unless either party gives at least thirty days' written notice before the end of the then-current term; the licensor also reserves the right not to offer renewal (clause 8). Invoicing is for a twelve-month period in advance unless the Product Terms provide otherwise, with applicable taxes added (clause 9).

Undisputed invoices unpaid after thirty days attract interest at eight per cent above the Bank of England base rate under UK late payment legislation, together with collection costs including agency and reasonable attorney's fees and court costs, and the licensor may suspend access to the Product without prior notice while payment is outstanding (clause 9). Product-specific commercial and operational provisions sit in separate Product Terms which are not published and which prevail over the general Terms in the event of a clash (clauses 1.0 and 17.0). The only published commercial route is a demonstration request.

Confidentiality and data terms: No business associate agreement or health-data tier is offered and none would be expected on a patent search platform. The data instruments that do exist are published and ungated, which is unusual in this lane: the UK and International Annual Subscription Terms are downloadable in full, and a Privacy Policy and Cookie Policy sit in the site footer. The Terms carry the substance a procurement reviewer looks for. Clause 19 imposes a mutual confidentiality obligation lasting the term plus five years, limits each party's use of the other's confidential information to performing the agreement, and requires notice before any legally compelled disclosure so far as the law permits. Clause 20 commits both parties to the UK Data Protection Act 2018 and the EU General Data Protection Regulation. Clause 13 gives the customer a vendor indemnity against third-party intellectual property claims arising from the Product. Clause 4(b) allows the customer to retain an archive copy of retrieved Information for twelve months after termination, subject to the clause 5.1 restrictions against searchable database storage and external redistribution. Clause 6 makes the customer responsible for User ID security and gives the licensor an audit right on seven days' notice. Two absences are recorded rather than inferred: no data processing addendum was located, and no subprocessor list is published, so the identity of any processor is not established.

Note: Commercial structure read from the UK and International Annual Subscription Terms, published as an ungated PDF and last modified June 2025, together with the product estate, on 13 September 2026. This row is written under R17 rather than R10's figure trigger: no price appears anywhere on the vendor's surfaces, but the agreement publishes the unit of charge and the billing structure, which lifts Commercial Transparency off the D floor and therefore obliges a row. entryPriceUsd is null because no figure is published at any level, following the convention that null records the absence of a published figure and is never conflated with zero. Currency is not fixed by anything published: the vendor is a UK company invoicing in an unstated currency, quotes UK, US and German telephone contacts, and the late-payment provision is expressed against the Bank of England base rate, which points to sterling as the contracting currency without establishing it. The Terms state at clause 17 that Product Terms prevail over the general Terms in the event of a clash, so the commercial specifics for PatBase itself sit in a document that is referenced and not published. No pricing page exists in the site navigation and the only published route to a figure is a demonstration request.

Legal Signals

What each signal means

A signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.

Confidentiality and Privilege

Client Data in Training

Can material a lawyer puts into this product be used to train a model?

Never, in policy only

A public policy or trust page states no training on customer content, with no matching term located in the published agreement.

Public material states plainly that customer content never trains models, with no matching term in the published agreement, which is this value. The statement appears on more than one product surface in the same words and is unhedged: the vendor never uses customer data to train models, and a customer's searches and results are their own, safeguarded within a private cloud. It sits beside a data protection commitment and a statement that information is stored securely and privately.

R43(1) was run and, unusually for this lane, can be discharged. The UK and International Annual Subscription Terms are published as an ungated document and were read in full. They contain no training provision in either direction: clause 3 vests intellectual property in the licensor and its suppliers, clause 4 grants the customer its licence to use and archive Information, clause 19 imposes mutual confidentiality with a five-year tail, and clause 20 commits both parties to the UK Data Protection Act 2018 and the GDPR.

Nothing addresses model training. So the prohibition is established as a policy commitment and is established as absent from the contract, which is the precise state this value records rather than an inference from an unread instrument. A buyer relying on it is relying on a published statement, not on a term they could enforce.

Source: Vendor PublishedWe never use your data to train models. Your searches and results are your ownAs of Sep 13, 2026Evidence

Prompt and Output Retention

How long does the product keep what a lawyer typed, and can that be set to zero?

Not addressed

No located public material states how long prompts and outputs are retained.

No located public material states how long the customer's searches, queries or generated answers are kept. The agreement was read in full and addresses retention only in the opposite direction, which is worth recording so the absence is not mistaken for silence about data generally. Clause 4(b) grants the customer a right to store an archive copy of Information obtained through the platform for up to twelve months after the agreement ends, and clause 5.1 restricts how that copy may be held, forbidding a searchable database format or external redistribution without written permission and additional fees.

So the terms regulate what the customer may keep of the vendor's data, and say nothing about what the vendor keeps of the customer's activity. On the product surfaces the position is a characterisation rather than a period: searches and results are stated to be the customer's own, held within a private cloud, with information stored securely and privately. Nothing states a retention period, a deletion right, or what becomes of search history at the end of a subscription.

The gap has practical weight because search history is not incidental here: the platform invites users to bulk-train alert classifiers from their folders and search history, so that history is retained long enough to be a training input and nothing says for how much longer.

Source: Vendor PublishedAs of Sep 13, 2026Evidence

Ethical Walls and Matter Segregation

Does retrieval respect the firm’s ethical walls, or can the model read across them?

Claimed, not documented

Segregation is asserted in public materials with no published detail on how it is enforced.

Separation is claimed and no permission model is documented, which is this value. What is claimed is real and comes from two places. On the product surfaces the platform is described as a private cloud in which a customer's searches and results are their own, stored securely and privately, which addresses separation between customers. In the agreement, access is organised around individually issued User IDs: clause 5(e) forbids a User ID being used by anyone not employed by the customer, and clause 6 makes the customer responsible for maintaining the security of its User IDs and requires it to notify the licensor immediately of any unauthorised usage, with a reciprocal audit right for the licensor on seven days' notice.

So identity is individuated and the customer carries a contractual duty around it. What is not documented is any model for controlling access between those identities. No roles are enumerated, no administrator capability is described, and nothing states whether one user's saved work, folders, search history or alerts are visible to colleagues on the same subscription. The concern is concrete for the buyers named: a patent firm running searches for clients competing in one technology field needs a wall, and a corporate team may need to shield an unfiled application from colleagues. Nothing published establishes whether that is achievable.

Source: Vendor PublishedAs of Sep 13, 2026Evidence

Third Party Request and Subpoena Notice

If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?

Notice committed

Terms commit to notice where lawfully permitted. No transparency report located.

Notice before compelled disclosure is committed in the published agreement, which is this value, and it is the only record in this lane where the commitment could be established at all. Clause 19 imposes a mutual confidentiality obligation, restricts each party's use of the other's confidential information to performing its rights and obligations, and then handles compulsion expressly rather than by silent exception: a party may disclose to the extent required by law or by any government or regulatory authority, provided that, to the extent it is legally permitted to do so, it gives the other party as much notice of the disclosure as possible.

The obligation runs for the term and five years afterwards. Two features are worth naming. The duty is mutual, so it binds the vendor in the customer's favour rather than merely preserving the vendor's position. And it is drafted to give the customer the practical benefit notice exists for, which is time to seek a protective order or narrow the demand. What is not present, and is recorded so the value is not read as stronger than it is: no commitment to challenge or resist a demand, no undertaking to seek to redirect the request to the customer, no minimum notice period, and no transparency report of any kind.

The notice duty is also framed by reference to confidential information generally rather than to customer data specifically.

Source: Vendor Publishedit gives the other party as much notice of the disclosure as possibleAs of Sep 13, 2026Evidence
Accuracy and Authority

Primary Law Corpus Provenance

Where does the law in this product come from, and does the vendor have the right to use it?

Sources named, basis unstated

Sources are identified without stating the licence or rights basis.

Sources are named at the level of issuing bodies with the basis on which they are held left unstated, which is this value. The naming is real and in one area is precise. The corpus is described as more than 180 million publications from over 100 patent issuing authorities, with full-text coverage of major jurisdictions delivered through a live feed of what the vendor calls quality patent sources, subject to daily quality checks and improvements to raw patent data.

For standard-essential patent work fourteen standard-setting organisations are named individually as integrated sources, being IEEE, ISO, CCSA, OMA, IMT, OneM2M, TIA, ATIS, ITU-T, ITU-R, IEC, IETF, ANSI and TTA, each searchable by a dedicated syntax with source and company columns exposed in the results table. Machine translation is attributed to a named partner, RWS. What is not stated is the footing. No licence, agreement or data supply arrangement is described for any source, no individual patent office is named among the hundred, and no coverage dates or backfile depth are published for any jurisdiction, so a searcher cannot establish how far back a given authority runs.

The agreement gestures at the position without resolving it, clause 10 recording that some features depend on public patent services provided by patent offices or other third parties and disclaiming responsibility for their quality or availability.

Source: Vendor PublishedAs of Sep 13, 2026Evidence

Good Law Verification

Does the product tell you when the authority it just cited has been overruled?

Licensed citator, surfaced inline

Treatment signals come from a named commercial citator and appear with the authority.

Third-party status data is carried and surfaced inline within the record, which is this value, with R15 governing the translation because on a patent platform the good-law question is whether a right remains in force. Legal status updates are described as part of what the platform provides alongside bibliographic data, full-text records and family information, so a searcher assessing a result can see its standing without leaving the platform, and family-level organisation means status can be read across a family rather than one publication at a time.

The data originates with the issuing authorities rather than with the vendor, which is what places this at the licensed rather than the own-signal tier: the agreement makes the dependency explicit, clause 10 recording that features are based on public patent services stored and provided by patent offices or other third parties, that changes to those services are outside the licensor's control, and that no responsibility is taken for their quality.

Two limits follow and are recorded. Nothing published states how frequently legal status is refreshed or how quickly an authority's change propagates, which on a freedom-to-operate question is the difference between a live blocking right and a lapsed one. And no vendor-generated status determination, prediction or confidence indicator is offered, so the platform reports what the offices report and adds no judgement of its own.

Source: Vendor PublishedAs of Sep 13, 2026Evidence

Refusal and Uncertainty Behaviour

What does the product do when the answer is not in the corpus?

Confidence signal only

The product exposes a confidence or grounding score without an explicit abstention path.

A confidence signal is published and no behaviour at the point of doubt is described, which is this value exactly. The signal is real and is a designed feature rather than an artefact: results carry relevance scores, described as helping a user prioritise, with the AI stated to learn what matters to that user. Alongside it AI Explain shows precisely which sections and wording of a document overlap the query, so a searcher can see the basis of a score rather than only its magnitude, and the vendor frames that as letting the user immediately judge how relevant a result is.

Taken together a user gets a graded output they can interrogate, which is more than a ranked list. What is absent is anything about the system's own uncertainty. No threshold is stated below which a result is withheld or flagged, nothing describes what Document Chat does when a question cannot be answered from the patent in front of it, nothing indicates when Smart Images has failed to resolve a reference numeral, and no state exists in which the platform reports that it does not know.

The failure mode that matters on a prior art tool is the relevant document that never appears, and a relevance score speaks only to what did appear. Nothing published tells a searcher when the ranking itself should not be trusted.

Source: Vendor PublishedAI learns what's important to you and helps you prioritise with relevance scoresAs of Sep 13, 2026Evidence

Fabricated Citation Record

Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?

None located

No court order, opinion or disciplinary record addressing fabricated or hallucinated legal citations produced by this product has been located as of the date shown. This is a statement about the public record on that one subject, not a finding about the product, and this signal is not a litigation history.

Searched on 13 September 2026 against the company name and the product name, across reporting and trackers covering court decisions on AI-generated fabricated citations. None located. No decision, sanction or disciplinary referral names Minesoft or PatBase. The absence was tested against directly comparable material, patent practice now being represented in the case law: in Lexos Media IP LLC v Overstock.com in the District of Kansas, counsel were ordered to show cause over briefs containing nonexistent quotations, nonexistent and incorrect citations and misrepresentations about cited authority, and were fined 12,000 dollars in total across four lawyers, the responsible attorney having filed a declaration admitting he authored the document and failed to check the AI-generated citations before filing.

The tool involved was a general-purpose chatbot. Under R119 this signal records fabricated legal citations in filings and nothing else, so no other proceeding involving this vendor would appear here. One point of product context: the platform retrieves patent documents from curated authority feeds rather than generating citations to legal authority, and Document Chat answers questions about a specific patent already in front of the user, so the exposure this signal tracks arises only indirectly.

Source: Bar Guidance or Court RecordAs of Sep 13, 2026
Professional Responsibility

Bar Guidance Alignment

Has the vendor engaged in public with the ethics opinions its buyers are bound by?

Not addressed

No located public material engages with bar or ethics guidance.

No located public material engages with bar, ethics or practitioner guidance. No bar association, rule of professional conduct, ethics opinion, patent office code of practitioner conduct or regulator is named or mapped to the product, in the United Kingdom, the United States, Germany or any other jurisdiction in which the vendor operates. Nor is professional responsibility engaged generically: the agreement contains no use condition asking the customer to operate the platform consistently with its own professional obligations, and the product surfaces contain no statement that the tools support rather than replace professional judgement.

The omission is squarely established rather than untested, the agreement having been read in full and the product, search, analytics and feature-release pages having been read. It is worth naming against this product's own positioning. The platform is sold for freedom-to-operate analysis, invalidity screening and patent due diligence to an audience the vendor expressly describes as including legal teams and R&D teams together, and patent practice carries its own professional regime, including the duty to disclose material prior art, which a prior art search tool touches directly.

Nothing on the estate connects the product to that duty or to any other. Verified against the surfaces read on the date shown.

Source: Vendor PublishedAs of Sep 13, 2026Evidence

Billing and Fee Posture

Does the vendor address what happens to the bill when the work takes an hour instead of six?

Not addressed

The product sits inside a lawyer to client fee relationship and no located public material addresses billing, fee or disclosure treatment, with no savings claim published either.

Nothing published addresses what happens to the bill when AI-assisted work takes an hour instead of six, which is the floor. The question arises squarely: the vendor sells to law firms and describes itself as helping patent professionals win back time, its AI features are marketed on speed and on reducing noise, and prior art searching is classically billed to a client either as practitioner time or as a disbursement.

Nothing follows in disclosure terms. No per-matter record of AI-assisted searching is described, nothing marks a result set or a Document Chat answer as machine-generated for the purposes of a search report or fee narrative, and no guidance is published on fee or disclosure treatment for a firm passing search cost to a client. Recorded and expressly not credited under R21 and R24 because they answer a different question: the agreement does publish the commercial mechanics between vendor and customer in detail, being per User ID licensing, twelve-month terms invoiced annually in advance, pro rata charging for User IDs added mid-term, and interest at eight per cent above base for late payment.

That is transparency about what the customer pays the vendor, which is graded on Commercial Transparency, and it says nothing about what the customer's own client is told or charged.

Source: Vendor PublishedAs of Sep 13, 2026Evidence

Outside Counsel Guideline Readiness

Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?

Not addressed

No located public material supports a client side disclosure obligation.

One of the three artifacts a firm would need is present and the two that matter for an AI clause are not, which leaves this at the floor. What exists is genuinely forwardable and should be credited in the summary even though it does not lift the value: the UK and International Annual Subscription Terms are published as an ungated document a firm can download and send to a client, and they contain the substance a client usually asks about, being mutual confidentiality with a five-year tail, a commitment to give notice before any legally compelled disclosure so far as the law permits, an undertaking to comply with the UK Data Protection Act 2018 and the GDPR, and a vendor indemnity against third-party intellectual property claims.

Very few vendors in this corpus publish that much. What is absent is what an AI clause actually asks. There is no subprocessor list of any kind, and no data processing addendum was located, clause 20 committing to data protection law without identifying any processor. And no model provider is named for any of the seven AI capabilities, so a firm asked which AI providers see its client's search terms or its questions to Document Chat cannot answer.

The one third party identified anywhere, RWS as machine translation partner, is named as a co-developer rather than disclosed as a processor of customer data.

Source: Vendor PublishedAs of Sep 13, 2026Evidence

Court Disclosure Support

If a judge’s standing order requires an AI disclosure, can the product produce one?

Not addressed

No located public material addresses court disclosure or verification certification.

No located public material offers a record of AI involvement intended to be shown to anyone outside the customer. What exists is transparency at the moment of use rather than an artifact. AI Explain shows the user which sections and wording of a document overlapped their query, which is real explainability and is graded on the citation accuracy and autonomy rows, but it is a view within the interface rather than a record that persists and travels.

Search history exists and is retained sufficiently to be used as training input for alert classifiers, and analytics output including charts, diagrams, heat maps and clusters can be exported and shared, so material does leave the platform. But nothing published frames any of it as evidence of how a search was conducted. No search report template is offered, nothing marks a result set as AI-expanded rather than Boolean-retrieved once exported, no export of the AI Explain overlap is described, and no guidance addresses whether or how the use of the AI features should be recorded or disclosed to a patent office, an opposing party in invalidity proceedings or a client questioning the adequacy of a search.

On a platform sold for freedom-to-operate and invalidity work, where the defensibility of the search is the thing later challenged, that is the gap.

Source: Vendor PublishedAs of Sep 13, 2026Evidence
Contact

Correct a record, or ask how something was graded

Every grade and every signal on this index is drawn from public sources and dated. If a record is wrong, out of date, or missing an artifact the index did not locate, send the source and it will be reviewed and the record redated. Vendors are welcome to submit documentation. Nothing on this index is for sale, including a listing, a placement, or a grade.

AI Legal Index

The AI Legal Index is an independent index that tracks changes to AI vendors in legal. It holds 61 vendors across 9 categories, each graded on the same 15 capability axes and recorded against 12 legal signals, from public sources. No vendor pays for inclusion, placement, or rating.

Index Status
Last index update
September 13, 2026
The AI Legal Index is an editorial reference. It is not a regulatory body, not a law firm, and nothing published here is legal advice or a recommendation to retain or avoid a vendor. Records are verified against published sources, bar guidance and public court records. Where a record reads not addressed, the material was not located in public sources on the date shown. See the Methodology page for evaluation standards and limitations.
© 2026 AI Legal Index
3801 N Capital of Texas Hwy, Ste E240 · Austin, TX 78746