P

PatentSight+

The models power named capabilities on a platform that works without them. Four AI capabilities are published. Protégé, a generative assistant that turns plain language business questions into structured visual answers, has been generally available since May 2026. AI-Powered Features explain charts and speed analysis, the AI Classifier builds custom technology views, and TechDiscovery, added in October 2024, allows patent search in ordinary words.

Underneath sits a platform with its own longer history: a harmonized database of more than 90 million patent family records, manually checked ownership data, legal status tracking, more than 100 attributes and measures including the Patent Asset Index, full text and field search, similarity search, saved search history, and the visualization tools the product was first known for. Remove every model and the analytics platform remains complete.

The vendor positions Protégé as changing who can use the platform rather than what it does, replacing complicated filters with simple questions so business leaders and others outside IP can reach the same data.

Grounding is real and documented, and there is no measured accuracy figure. The grounding goes beyond citation to reproducibility. Every Protégé answer comes with the full search query the assistant built, stated to be fully reproducible within PatentSight+, so a user can rerun the analysis and get the same result. The assistant also explains each step, shows its queries, puts results in context and suggests next steps.

Answers rest on the curated harmonized dataset and on published metrics, including the Patent Asset Index method, so the logic can be inspected as well as the data. The assistant's own accuracy is not measured. No precision, recall or error rate is published for Protégé, no test set is described, and no failure mode is named. The figures the vendor does publish measure effort, not correctness, and come from users: reductions in manual analysis effort of up to 70 to 90 percent and up to three times more output.

An on demand session is advertised on independently validated accuracy for AI patent classification, but no result is published.

A written commitment that the model works alongside the user is backed by specific review features, but not by a full control structure. Protégé is said to keep professionals in the driver's seat, and human oversight is one of the five RELX principles the product page says govern all its AI. The review features are concrete. The assistant explains every step of its analysis, shows the full queries it built, puts results in context and suggests next steps, and each answer carries its query so the user can reproduce and inspect it.

Support documentation gives guidance on prompting and on validating AI generated insights. Vendor material quotes a named customer valuing the chance to see how the assistant reasons. Missing are any threshold at which a capability acts without review, any distinction between modes with different levels of automation, any confidence signal on an answer, and any account of what happens when an output is wrong or how to report it.

Nothing is published about the AI Classifier or the enrichment models, whose outputs enter the dataset rather than arriving as answers a user inspects.

Named customers say specific things, but no measured figure is tied to them. Siemens describes using the platform to track patent quality over time against competitors, accounting for acquisitions and divestitures. It notes that the quality indices reflect differences between the US, German and Chinese markets. Christopher Hauke, Head of Strategic IP and Innovation at Schott Pharma AG and Co. KGaA, values seeing how Protégé reasons through a problem, saying it makes the analysis transparent and gives him confidence in the results.

The AI Insider Program is said to have developed Protégé with hundreds of participating organizations, with early access feedback shaping its reasoning. The figures are the weak part. Reductions in manual analysis effort of up to 70 to 90 percent and up to three times more output are attributed to users, not measured by the vendor. They are given as ranges with an upper bound and carry no method, baseline or sample. No figure is tied to either named customer.

Published commitments are substantive on most points and fall short on two. The product page excludes training in express terms: queries and usage data are never used to train AI models, and customer data is never used to train AI models. Data is not shared outside the customer's organization. A commitment across products is also published: data entered into Protégé is not shared with other LexisNexis products unless explicitly communicated and authorized, which matters because the vendor runs seven adjacent IP products a buyer might assume share data.

Retention is addressed as policy and as a control: the vendor cites retention and deletion policies, and users can turn off automatic saving of queries and compare, combine, export or delete saved ones. Personal data is handled under a published Privacy Policy and Data Processing Addendum, and the General Terms return Subscriber Files on request within sixty days of termination. Privilege and work product are not addressed by name.

No model provider is named, so nothing says what a third party sees of a query, which matters for strategic questions about unannounced acquisitions and licensing positions.

A contractual scope limit sits in the terms, while the marketing speaks in decision terms and widens the audience. The General Terms grant a nonexclusive, nontransferable, limited right to use the online services and materials for research purposes, which frames the product as research rather than advice, and the terms address legal professionals separately. The marketing, though, speaks of decision ready insights and high value strategic decisions, with worked examples that are professional questions, such as identifying top acquisition targets in a sector and finding licensing partners for a technology.

The vendor openly positions Protégé as widening access to patent data across the organization and letting business leaders reach insights without complex filters. So people who are not lawyers are an intended audience for output that feeds licensing, acquisition and portfolio decisions. Nothing published says what a business user should not do with a Protégé answer, or where a patent attorney's judgment must come in.

A published governance framework, defined outside the product and expressly applied to it, without testing results or a named owner. The framework is the RELX Responsible AI Principles. The product's AI page says all AI in PatentSight+ is governed by them, and the TechDiscovery release says that module was built in line with them. The five principles are listed on the product page: real world impact, minimizing bias, transparency in how insights are generated, human oversight, and governance with strong privacy and security practices.

A Responsible AI page on the IP site argues the case for self regulation rather than only claiming compliance. Structured feedback exists: the AI Insider Program gives participating organizations early access, focus groups and feedback channels to surface problems before general release. No person, committee or function is named as accountable for model behavior, and no testing regime before release is described. Bias minimization is named as a principle, and an on demand session advertises independently validated classification accuracy, but no evaluation result is published.

Published policy covers most of the ground. Training on customer data is excluded in express terms, and repeatedly. Encryption is stated for data at rest and in transit, with privacy by design named as a development practice. LexisNexis cloud environments, products and security programs are audited annually, and the audit scope is listed: encryption, backup, disaster recovery, access controls, data destruction, breach prevention and confidentiality safeguards.

Incident practice is published and tiered, with technical, administrative, business and executive escalation. Subprocessors are handled well. The Data Processing Addendum points to a maintained public subprocessor list, commits to updating it at least fourteen days before any change, and gives the customer fourteen days to object with reasons. Retention and deletion policies are said to exist, and users control their saved search queries directly.

Two gaps remain. No retention period is stated anywhere. Access control at product level is thin: single sign on across LexisNexis applications is described as being introduced, not available, and no role or permission model is published.

A published agreement with a real indemnity, and AI output left to the user to verify. The LexisNexis General Terms and Conditions (effective 5 June 2026), which the product links through the LexisNexis terms hub, have LexisNexis defend and indemnify the subscriber against third-party patent, copyright, trademark and trade secret infringement claims arising from use of the service, with that indemnity excluded from the twelve-month liability cap.

The AI Services clause says AI output may not be accurate and users must verify it, so nothing stands behind a wrong AI insight. The terms also give 30 days to cure a change that materially degrades the service, failing which the subscriber may terminate, and return of subscriber files on request within 60 days of termination.

Integration is claimed and partly named, without a published catalog or configuration detail. Most of what is established sits inside the LexisNexis family. Single sign on across LexisNexis applications is described as letting users move between products without logging in again, but the vendor says it is being released over time rather than available now. Data can leave the platform: search queries and results can be exported, and a separate product, IP DataDirect, provides a data feed; an older factsheet mentions a developer portal and proof of concept access.

TechDiscovery and the AI Classifier are modules within the platform, not integrations. No connector catalog is published, no external system is named as supported, no API documentation for this product was reached, and nothing describes the direction of data flow or the setup required. That matters for a corporate IP department that runs a docketing or IP management system alongside its analytics, since nothing says whether portfolio data can move between them.

Cloud delivery is evident, and neither the tenancy model nor the region is stated. The product is a browser based subscription platform reached through LexisNexis online services, and the General Terms define it as access to online services rather than installed software. Hosting is described only in general terms: the vendor partners with trusted, enterprise grade cloud providers and monitors and improves its infrastructure.

That describes a class of provider without naming one or giving a location. Nothing names a data center region, a country of processing or an option to choose one, a real gap for a platform sold across Europe and Asia to customers with residency obligations, and an odd one beside the detailed data protection material the same vendor publishes. Tenancy is not addressed: nothing says whether saved searches, exports and Protégé history sit in a shared or isolated environment, and no single tenant or private option is offered or refused. The annual audit covers backup and disaster recovery, which is resilience practice, not a residency statement.

Certification is mentioned without a named standard held for this product, and the one standard named is promised rather than held. The product support documentation says Protégé in PatentSight+ follows the same enterprise security framework used across LexisNexis products and will be ISO 27001 certified. That is future tense: as of this check, the product is not certified on its own account. The vendor says it undergoes regular third party audits to maintain industry leading certifications, without naming any.

LexisNexis cloud environments, products and security programs are audited annually, with the scope listed: encryption, backup, disaster recovery, access controls, data destruction, breach prevention and confidentiality safeguards. Audit practice and its coverage are described, but the resulting attestations are not named, dated, scoped or obtainable. No trust center with reports is published for this product.

The vendor describes purpose built AI without saying what sits underneath. Protégé is described as purpose built for patent intelligence, as going beyond what general large language models and AI powered IP workflow tools can provide, and as having agentic reasoning refined through early access feedback. TechDiscovery, the AI Classifier and the enrichment models are each described by function. None is attributed to a model, version, architecture or provider.

The comparison with general large language models suggests the assistant is more than a wrapper without saying what it is, and nothing separates models built in house from models licensed and fine tuned. Nothing says where inference runs, what any provider keeps of a query, or whether a customer would be told if the model changed. The data is described in great detail, down to 90 million patent families, ownership data and more than 100 named attributes, while the models reading it are described only by purpose.

The reference to trusted, enterprise grade cloud providers names a class of infrastructure supplier, not a model supplier.

No pricing information is published at any level, including the unit of charge. The product pages, AI pages, Responsible AI page, resources and news, support centers and the LexisNexis legal and terms pages were all checked. There is no pricing page and no figure, band or unit on the vendor's own pages; the routes offered are a demo request and a contact form. The terms confirm that pricing is negotiated rather than public.

The Subscription Agreement is defined as the General Terms plus Supplemental Terms plus the rates in the Price Schedule, and the terms distinguish transactional pricing plans from fixed price and fixed term arrangements without publishing either. A buyer can learn that two charging shapes exist, and nothing about what either costs. Third party listings describe customizable pricing by quote and subscription term pricing, which confirms the absence rather than adding structure.

Coverage is described with substance across data, buyers and uses, and the boundaries are left open. Data coverage is quantified: more than 90 million patent family records, harmonized and enriched with manually checked ownership data, legal status tracking and more than 100 attributes and measures. Buyers named range widely: corporations, law firms, regulators and academic users. The vendor says the assistant is meant to widen access from IP specialists to business leaders and others in the organization.

Uses are listed concretely, with worked example questions on acquisition targets in a named sector, how a named technology landscape has changed, finding licensing partners and cutting portfolio fees. Benchmarking, due diligence, competitive analysis, litigation risk and identifying nonpracticing entities are also listed. A named customer describes quality indices that account for differences between the US, German and Chinese markets.

No limits are stated. No technology area is named as better or worse covered, no jurisdiction is out of scope, nothing says what the AI Classifier handles poorly, and nothing separates this platform's coverage from the adjacent LexisNexis IP products.