D

Doctrine

The models drive the search and drafting layer and sit on a product that predates them. Underneath is what the vendor describes as 80 million interconnected French legal documents, searchable by keyword, court, company, advocate, date and decision number, which was the product for years before generative features arrived. The model work on top includes a legal chatbot answering in natural language from the legal corpus, a documents chatbot answering only from files the user supplies, and a company chatbot over corporate and litigation records.

It also includes decision and instrument summarization, extraction, comparison, conformity checking and translation. Flow Litigate assembles classification, chronology, statement of facts and bordereau for a contentious file, and Flow Counsel analyzes contracts for risk. Strip the models out and a large searchable legal database remains.

Grounding is real, documented and specific about its boundaries, without any published measurement. The vendor's own blog states that the Chatbot Juridique answers exclusively on the basis of the legal corpus, while the Chatbot Documents relies only on the documents supplied, and that both always cite their sources. It sets out the reasoning too: a professional legal tool must let the user verify the origin of every assertion, and the way to get the best from these systems is to confine them to precise legal sources they can cite.

That is a described retrieval method plus linked primary sources. The vendor also publishes, on its marketing blog, an analysis by Damien Charlotin, the researcher behind the AI Hallucination Cases database, of the first French decision involving AI hallucinations. And it argues that general purpose AI is unsuitable for legal work because a fabricated authority or a stale rule has concrete consequences. No accuracy figure, test set, evaluation or error rate is published.

Autonomy is claimed across substantial work, and the only oversight offered is source citation. What the product does unattended is described plainly. Flow Litigate takes a contentious file and produces classification, a chronology, a statement of facts and a bordereau automatically. Flow Counsel analyzes contracts, identifies risks and returns what the vendor calls legal recommendations in a few clicks, and the company chatbot condenses each corporate instrument into a summary described as clear and exhaustive.

Oversight rests on verifiable sources: both chatbots cite what they relied on, and the vendor's stated position is that the user must be able to check the origin of every assertion. That is a real review point. No control structure is published: nothing states what a person should check before a generated chronology or bordereau is used, sets a threshold at which the system defers, or covers what happens when a summary or a risk recommendation is wrong.

Real deployment evidence with substance, short of figures for what changed. The scale is first party and recent: the RELX completion release of 3 September 2026 states that Doctrine is used every day by 27,000 legal professionals across France, Italy, Germany, Spain and Luxembourg, and describes its customers as law firms of all sizes, large multinational corporations and public sector organizations including French government ministries, local authorities and universities.

That is a named sector spread rather than a logo wall. An institutional partnership with the Paris Bar gives members preferential subscription terms, which is a third party with its own reputation choosing to stand behind the product. A named advocate in company law is quoted on the record, with a linked interview, saying he uses the platform systematically in both advisory and contentious work to get a precise and exhaustive picture of a company's position.

What is missing is the quantified half: no figure for time saved, matters handled or outcomes changed, and no method behind any claim.

Substantive published commitments on confidentiality and training, short of documented segregation. The security FAQ for documents states that documents and data are never used to train Doctrine's algorithms or AI models, that no Doctrine employee can access files except developers and product staff correcting an error or studying product use, that files go to no third party beyond the subprocessors the service needs, that documents pass through Microsoft Azure's European servers and are deleted immediately, and that Google and Microsoft are instructed not to retain them.

The vendor blog adds strict isolation between customers and hosting only in the European Union, and its Document Analyzer post (2020, updated May 2026) says a confidentiality agreement in the CGV protects the customer's secret professionnel. Isolation between customers is asserted without detail, segregation between matters within a firm is not addressed, and the access exception extends to studying product use.

What is published describes the product in advice terms while nothing addresses where its output stops and a lawyer's judgment begins. Flow Counsel is described as analyzing contracts, identifying risks and returning legal recommendations; the legal chatbot answers substantive questions of French law; and the marketing addresses advocates and in house counsel directly. Against that, the vendor's published position on verification is real: it argues that a professional tool must let the user check the origin of every assertion, which places responsibility with the user.

What is absent from the product pages is any statement that an output is not legal advice, that the subscribing lawyer remains responsible for the determination, or what the tool does not cover. The audience is unambiguous and professional.

Responsible AI positions are published without a mechanism a buyer could audit. The vendor writes publicly about the failure modes of legal AI. It argues that hallucinations have concrete consequences for a strategy or a file, that general purpose models are unsuitable for professional legal work, that reliability depends on the quality of the sources a system draws on, and that context management is key to useful output.

It publishes an outside researcher's analysis of the first French hallucination decision on its own blog. That is a stated position on how legal AI should behave, not governance of this vendor's own models. Nothing names who is accountable for the chatbots' behavior, describes what is evaluated before a change ships, or publishes any result. Nothing discloses whether output quality is uneven across areas of law, jurisdictions or the five languages the platform serves. No responsible AI policy, model card or bias statement is published.

Substantive published practice covering most of the ground, short of a subprocessor list a buyer can read and any incident practice. Retention: the security FAQ for documents says documents are held only for the time the provider needs to return its analysis, are deleted immediately from the Azure servers they pass through, and that Google and Microsoft are instructed not to keep them; the vendor blog lists deletion of customer data as available.

Access control: no Doctrine employee can access files, save developers and product staff correcting an error or studying product use. Encryption of files and transfers and strict isolation between customers are stated on the blog, with hosting on AWS in Frankfurt. Subprocessors: the AI providers are named (OpenAI GPT models through Microsoft Azure, and Gemini through Google Cloud, on European servers), and a subprocessor list for upload features exists in the vendor's personal data guide and was not read.

No incident notification practice is published; the blog says only that no security incident has occurred. The Trust Center at trust.doctrine.com, on Vanta, was not read.

The allocation of risk is published in French, accepting responsibility in one area and excluding it almost everywhere else. Article 7.1 of the general terms of sale (conditions générales de vente, in force from 23 January 2026) commits Doctrine to strict confidentiality over documents and data. That includes a ban on training or fine tuning any AI model on them and notice within two working days of any foreign authority's request for access, and breach may engage Doctrine's liability.

Article 7.3 commits Doctrine to carry professional liability insurance, without stating the amount. Article 10.3 warrants that Doctrine holds the rights needed to provide the services and guarantees their peaceful enjoyment, the French law counterpart of an intellectual property warranty. The same article says Doctrine cannot be held civilly liable for any damage arising from use of the document services, and article 9 excludes liability when generative AI features are unavailable because of its AI providers.

The site terms limit any liability to proven direct damage and say AI output never replaces a professional's analysis and must always be checked. No cap figure, indemnity or remedy for wrong output is stated.

Real integrations into the systems the work already lives in, named and briefly described, short of documentation. Two are published. The platform connects to a firm's own databases and document management systems, which the vendor frames as activating the knowledge already stored in the GED and limiting interruptions to the work, so a user's own precedents and files are searchable alongside the public corpus. And Doctrine runs inside Microsoft Word, described as enriching documents without leaving the working environment, which is where drafting actually happens for the buyer.

Those are connections into the right two places for a legal research and drafting product rather than a list of unrelated logos. What is not published is depth: no named document management systems, no statement of direction or sync behavior, no configuration requirements, and no implementer documentation. The integrations page itself was not read.

Residency and processing location are stated, short of deployment options or a documented tenancy model. The vendor blog says data is hosted only in the European Union, on AWS servers in Frankfurt, and the security FAQ for documents says the AI models that analyze documents run on Microsoft Azure and Google Cloud servers in Europe and that data never leaves the European Union, so processing is stated separately from storage.

The platform is delivered as a single web service; no alternative deployment, regional choice or tier difference is described, and tenancy is asserted as strict isolation between customers without detail. The Trust Center at trust.doctrine.com, on Vanta, was not read and may carry more.

A named certification with a date, short of published scope or evidence. The vendor blog (published 14 January 2025, updated 7 May 2026) states that Doctrine has been ISO 27001 certified since January 2025 after several months of verification by an independent auditor, that the certification is managed through Vanta, and that its hosting provider AWS holds ISO 27001, 27017, 27018 and 9001 certification. The certificate scope, the auditor and the report are not on any page read.

The Trust Center the blog points to, trust.doctrine.com on Vanta, is the place that would carry the scope and report, and its contents were not read.

The providers are named and where they run is stated, without any commitment to notify customers of changes. The security FAQ for documents says uploaded documents are analyzed with AI models hosted in Europe: OpenAI's GPT models supplied through Microsoft Azure on European servers, and Gemini through Google Cloud on European servers, with data never leaving the European Union. The vendor blog on adopting generative AI says users' questions are sent to a version of OpenAI's GPT models hosted only in Europe and that all of its providers are based in Europe.

The legal chatbot answers from the legal corpus and the documents chatbot from the user's own files, both with sources. No model version is named and no notice of changes is committed. The subprocessor list for upload features and the Trust Center at trust.doctrine.com were not read.

The shape of the offer is published, and no figure is. The pricing page presents three buyer profiles rather than a single enterprise funnel: the individual advocate, the law firm and the in house legal department. It answers the question that matters most about a research subscription: each subscription gives unlimited access to all Doctrine Pro features across every area of law, without segmentation, so a buyer does not pay per practice area or per module.

A free trial is offered. The Paris Bar partnership publishes a commercial term rather than a rate. Members get several free months and a discount over a twelve month commitment, with a second license free in the first year for firms of two advocates. No price, price range or unit of charge is published on the vendor's own pages.

Coverage is described with substance across buyers, jurisdictions and subject matter, and the boundaries are left open. Buyer segments are the structure of the offer itself, with distinct propositions for the individual advocate, the law firm and the in house legal department. The RELX release adds public sector bodies, including French government ministries, local authorities and universities. Jurisdictional reach is stated as France, Italy, Germany, Spain and Luxembourg, which the vendor frames as a global view with local impact.

A subscription reaches every area of French law without segmentation, with a tax encyclopedia called out separately, and the legal database is quantified at 80 million interconnected documents. No limit is stated: no area of law, jurisdiction or matter type is named as unsupported, no depth comparison is offered between the five countries, and nothing describes what the platform does not cover.