Advocacy is a litigation workspace built around case memory. Its premise is that a general legal AI answers each question from scratch, while a litigation team needs a system that already knows the matter. The platform maintains what it calls Axiom, a proprietary architecture that accumulates the facts, timelines, entities and contested issues of a case as it develops, rather than querying a generic legal database. Each query, review and note feeds that memory, so an associate who joins partway through discovery inherits the matter rather than rebuilding it.
Three surfaces sit on top. Associate is the conversational assistant that answers questions about the case. The AI Editor is a drafting environment native to Microsoft Word, with research beside the draft, generation streaming onto the page and citations verified. Extract pulls structured data from documents at scale. Version 2.0, released in December 2025, carried case memory across all three surfaces, and the company has said shareable case memory across a team comes next.
It is sold to midsize and Am Law litigation practices. Former securities litigator Téo Doremus is chief executive, and the company is based in San Francisco.
Side by side on the same 15 capability axes and 12 legal signals: where the two part company, and what neither one publishes.
Vendors in the same categories as Advocacy whose graded profiles look most alike, with a line on what each product does.
Similar is not the same, so check that each one does the job you need. No vendor pays for placement.
Ordered by how closely the graded profiles match across the 15 capability axes, inside the same categories. See how each axis is graded in the methodology.
All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.
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.
Remove the models and nothing remains to sell. The product is defined by its model architecture. Axiom is described as a proprietary case memory layer that ingests every document in a case, builds memory that persists across every interaction, and reasons across the entire matter rather than one document at a time. The vendor contrasts this expressly with querying a generic legal database. All three surfaces on top of it are generative: a conversational assistant that answers questions about the matter, a drafting environment with generation streaming onto the page, and a structured extraction tool.
The vendor rejects the sidebar plugin model on the ground that a plugin is bolted onto someone else's product, while this is the product. There is no document management system, matter management layer or review platform underneath that a firm would license on its own. The site blocks automated access, so this rests on material read through search results, and nothing here depends on a page that could not be opened.
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.
Accuracy is asserted and the mechanism behind it is named without being described. The material claims outputs a buyer can trust and states that citations are verified inside the drafting surface. It argues that grounding comes from case memory built on the matter's own record rather than a generic database, which the company frames as the system learning about the case before it does the work. So accuracy and the source of grounding are both addressed.
What is absent is any test or description: no accuracy figure, evaluation, test set or error rate, and no statement of failure modes. Nothing describes what verifying a citation consists of, or whether a user can open the source behind an assertion. The site blocks automated access, and any accuracy or benchmark page it publishes was not opened and may say more.
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 division of labor is implied by the shape of the product and never described as a control. The material presents a workspace the litigator works inside. Research sits beside the draft, generation streams onto the page the user is writing, and the recorded activity types are queries, reviews and notes attributed to people against docket entries. That describes a lawyer doing the work with the system assisting. The company also states that the system learns about the case before it does the work, which is about sequencing rather than supervision.
Nothing describes a review gate, an approval step, a threshold at which the system acts without a person, what happens when an output is wrong, or any abstention or confidence behavior. One line runs the other way, promising that what you draft is what you file. Oversight is implicit in the described workflow rather than absent from it. The product pages that would usually describe controls could not be opened and may say more.
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.
No production evidence is published in any material that could be read. The buyer segment is stated as midsize and Am Law firms, and no firm within it is named. No customer, case study, logo, deployment or figure appears in the home page material read through search results. None appears in the vendor's own December 2025 announcement either, which quotes only its own chief executive and chief operating officer. The interface material uses a fictional matter, Estrada v. Kincade, with invented docket entries, so nothing there is a deployment.
This rests on the announcement rather than on the blocked site, since a press release naming no customer is evidence in its own right. The customer pages of the site could not be opened and may say more. A named reference with a date or a figure would change this.
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.
The commitment that matters cannot be read before signing, and the vendor's own policy shows this. The published privacy policy states that the company does not use content customers upload to train its AI systems, which is a specific undertaking. The same policy scopes itself to the Website and says the handling of data uploaded to the AI platform is covered separately in individual Customer Agreements. A commitment in a document that excludes the platform has not been made about the platform.
The document that governs the platform is negotiated per customer and unpublished, so the commitment lives in a sales conversation and cannot be read in advance. Nothing addresses privilege or work product treatment, separation between matters or between firms, or what any model provider retains. That matters on a platform the vendor describes as holding strategy notes, evidentiary rulings and work product, and intends to make shareable across a team. Any security or terms page the site publishes could not be opened and may say more.
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.
Nothing that could be read addresses the advice line, on a product that drafts filings. There is no statement that the company is not a law firm, no disclaimer that output is not legal advice, and no description of the judgment needed before a draft is filed. Nothing addresses jurisdiction limits, supervision or competence. That holds across the home page material, the privacy policy and the vendor's own announcement.
The exposure is specific. The product is sold to surface critical facts and draft winning motions, its drafting surface promises that what you draft is what you file, and its own framing is that lawyers who control the context win. The terms of service, where such a statement would usually sit, did not appear in search results. That is weak evidence of absence rather than proof, and this note does not assert that the document does not exist.
The site blocks automated access; the home page and privacy policy were tried on 5 September 2026. Its terms and product pages were not opened and may say more.
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.
No governance position appears in any material that could be read. Nothing names an accountable owner or function for model behavior, describes evaluation before release, sets out principles or a policy, or claims a management system or certification, and nothing addresses uneven output. The gap has a specific shape here. A case memory layer that accumulates contested issues, strategy notes and evidentiary rulings across a matter, and that the company intends to make shareable across a team, is making judgments about what counts as critical context.
Nothing that could be read describes how that selection is governed or tested. The privacy policy is the one policy document that could be read, and it addresses website data handling without touching model governance. The site blocks automated access; the home page and privacy policy were tried on 5 September 2026. Any responsible AI or governance page it publishes was not opened and may say more.
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.
Nothing that could be read addresses retention, deletion or access for the material the platform holds, and the reason comes from a readable document. The privacy policy describes secure storage with a database provider, industry standard security protocols and regular security reviews. It retains personal data for as long as reasonably necessary, with no period stated, and names service providers by category only: a database provider, an analytics provider and payment processing.
It then scopes all of that to the Website and refers the handling of data uploaded to the AI platform to individual Customer Agreements. The stewardship material that exists covers website visitors, and the material governing a firm's case files is unpublished. No retention period, deletion commitment, access control model, encryption specification, subprocessor list or incident practice for platform data is published, and no AI or model provider is named anywhere.
No security or trust page appeared in search results, which is weak evidence of absence rather than proof. The site blocks automated access, tried on 5 September 2026, and any such page was not opened and may say more.
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.
Nothing published addresses who bears the loss when the system is wrong, and that follows from how the vendor contracts. The privacy policy states that the handling of data uploaded to the AI platform is covered separately in individual Customer Agreements. The governing document is therefore negotiated per customer and not published. No indemnity, liability cap, warranty position, disclaimer of warranties, service level commitment or insurance statement can be read before a buyer enters a sales process, and none appears in any material that could be read.
The exposure is real on a product whose drafting surface promises that what you draft is what you file. This rests on the vendor's own statement about where platform terms live, not on the blocked site. A published standard form agreement, if one exists, may say more.
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.
Connection is claimed by category and no counterparty is named. The vendor states that the platform synthesizes context about the case, including from ediscovery and legal research platforms. That identifies the two classes of system a litigation team runs and names no product or vendor within either. The drafting environment is described as native to Microsoft Word and as rendering.docx files with full fidelity. That is a file format and editing claim rather than an integration with a document or matter management system, and the company positions it against Word plugins.
Nothing describes what data moves between the platform and any external system, in which direction, on what trigger, or what a firm must configure, and no API or developer documentation appeared. Real classes of connection are named rather than shown as logos. The integrations material on the site could not be opened and may say more.
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.
Cloud delivery is implied throughout, and neither tenancy nor region is stated. The product is described as a shared workspace a team logs into, with case memory that persists across interactions and is meant to become shareable across a firm. The privacy policy refers to secure storage with a database provider, so delivery over the internet on infrastructure the vendor arranges is not in doubt. Beyond that nothing is available.
No statement says whether the platform is single or multi tenant, no dedicated or isolated option is described, no region is named for storage or processing, no cloud provider is identified, and no residency commitment is published. The only geographic fact is corporate: a San Francisco base, which locates the company rather than the data. No on premises or self hosted option is mentioned. Any security or enterprise page the site publishes could not be opened and may say more.
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.
No independent security attestation appears in any material that could be read. No SOC 2 of either type, ISO certification, penetration test summary, named auditor, examination period or other certification claim is published, and no trust center or portal appeared in search results. What the privacy policy offers is internal rather than independent: secure storage with a database provider, industry standard security protocols and regular security reviews.
No assessor, scope, date or report is attached to any of it. No badge or certification mark is published either. This rests on a documented absence in the material that could be read. The site blocks automated access; the home page and privacy policy were tried on 5 September 2026. Any certification or trust page the site publishes was not opened and may say more.
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 names its own architecture and identifies nothing underneath it. Axiom is presented as a proprietary case memory architecture, and the platform as context first and AI native, with three named generative surfaces built on it. A distinct architecture is therefore described rather than merely labeled. No model is named, and no version, provider or hosting arrangement, and no commitment to notify customers of a change is published.
The privacy policy comes closest and stops short. It names service providers by category only, as a database provider for storage, an analytics provider and payment processing. It identifies no AI or model provider at all, on a platform whose entire value is model driven. The architecture is described, but a buyer cannot learn from any readable source whose model reads a firm's matter file. The site blocks automated access, tried on 5 September 2026, and any model or subprocessor page was not opened and may say more.
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.
No pricing information appears in any material that could be read, at any level, including the unit of charge. No rate, tier, term or implementation cost is published in the home page material read through search results, the vendor's own announcement or the privacy policy, and no pricing page appeared in search results. Nothing states whether the platform is charged per seat, per matter, per firm or on usage. One documented fact points the same way.
The privacy policy records that platform data is governed by individual Customer Agreements, which indicates a negotiated commercial relationship rather than published rates. That fits the stated buyer segment of midsize and Am Law firms. Any pricing page the site publishes could not be opened and may say more.
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 claimed at a level of generality that does not tell a buyer whether their practice is served. The segment is stated with reasonable precision: midsize and Am Law firms running high stakes litigation. The scope within a matter is complaint through trial, spanning parties, claims, timelines, procedural history, evidentiary rulings, strategy notes and work product. That describes a stage of work rather than a practice.
No practice area is named. Nothing says whether the product suits commercial, employment, securities, product liability or class work. The only indirect signals are a fictional demo matter involving price fixing and market allocation, and a leadership team drawn from securities and antitrust practice. No jurisdiction is stated, no firm size range beyond the segment, no unsuitable matter type and nothing on languages. Any solutions or industries pages the site publishes could not be opened and may say more.
The public pages on file for Advocacy, 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.
Client Data in Training
Read Sep 5, 2026
Fabricated Citation Record
Read Sep 5, 2026
Alongside Promptless, Advocacy introduced a new pricing model. Firms open a matter for a nominal fee and then pay a la carte only for the AI generated contributions they accept into that matter.
Advocacy released Promptless, which proposes litigation tasks on its own when a document production arrives, opposing counsel sends a letter or a deadline approaches, drawing on connected documents, email, calendars and dockets. It classifies incoming documents and flags contradictions with the case theory, runs multistep research including EDGAR, drafts outlines and edits, and calculates deadlines from docket entries showing the governing rule and arithmetic. Every proposed action stays unconfirmed until an attorney accepts it, and the platform keeps an audit trail of human and AI contributions tied to the team's approved reading of each document. Promptless is model agnostic, so firms can bring their own model, and teams set per matter and per client which work the AI may do, where approval is required, which sections stay lawyer only and what information the AI may see.
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.
Can material a lawyer puts into this product be used to train a model?
A public policy or trust page states no training on customer content, with no matching term located in the published agreement.
A clear commitment in a policy, with a gap in scope that a buyer needs to see. The published privacy policy states that the company does not sell any personal information and does not use content customers upload to train its AI systems. On its face that is specific and unqualified. The same document opens by scoping itself to information collected through the website. It states that the handling of data uploaded to the AI platform is covered separately in individual Customer Agreements.
A policy that governs the website has not made that statement about the platform, and the document that governs the platform is negotiated per customer and unpublished. A search for a published customer agreement found none, which is weak evidence of absence rather than proof. No opt out or configuration setting is described. The site blocks automated access; the home page and privacy policy were tried on 5 September 2026, and a customer agreement, if published, may say more.
How long does the product keep what a lawyer typed, and can that be set to zero?
No located public material states how long prompts and outputs are retained.
How long the platform keeps anything it handles is unstated. The privacy policy addresses retention once, saying personal data is kept for as long as reasonably necessary to fulfill the purposes described or as required by law, with no period. That provision governs website information rather than platform content, since the same policy refers the handling of uploaded data to individual Customer Agreements. Nothing addresses how long documents, extracted facts, case memory, prompts or generated drafts persist, and no commitment to delete or return matter material was located.
Retention sits close to the core of this product. Case memory is designed to accumulate and persist across a matter and across staff turnover, with the company stating that knowledge never walks out the door. The site blocks automated access, tried on 5 September 2026, and any terms, security or trust page it publishes was not opened and may say more.
Does retrieval respect the firm’s ethical walls, or can the model read across them?
No located public material addresses walls or matter level segregation.
Walls and separation go unaddressed, and the product's own roadmap makes the question pressing. The platform keeps a distinct case memory per matter. The company has said it is deploying shareable case memory so that context compounds across a team, with new team members onboarding instantly into an existing matter. That deliberately widens who sees a matter's accumulated context, and nothing published describes the boundary around it.
There is no tenancy model, no permission or role model, no access control at matter level, and nothing on screened matters or a lateral hire arriving with conflicts. Nothing addresses separation between firms either. The privacy policy, the only policy document that could be read, governs website data and does not touch the question. The site blocks automated access and was tried on 5 September 2026; any security or product documentation page was not opened and may say more.
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
No located term or policy addresses third party requests for customer data.
No term or policy that could be read addresses third party requests for customer data. The privacy policy governs website information rather than platform content. It mentions using information to meet legal requirements but does not say what happens when a subpoena, court order or government demand reaches material held in the platform. It commits to no notice, reserves no discretion over notice and publishes no transparency report.
This evidence normally sits in the confidentiality section of a master agreement, and none is published: the privacy policy records that platform data is governed by individual Customer Agreements, negotiated per customer. The gap is pointed on a product holding the accumulated work product and strategy notes of contested matters, where a demand for the case memory itself is foreseeable. The site blocks automated access; the home page and privacy policy were tried on 5 September 2026. Any standard form agreement it publishes was not opened and may say more.
Where does the law in this product come from, and does the vendor have the right to use it?
No located public material identifies the corpus behind the product’s answers.
No corpus stands behind the product's output, and the vendor says so as a design claim rather than leaving it to inference. Axiom is described as building a dynamic, evolving intelligence layer that accumulates the facts, timelines, entities and contested issues of the matter as it develops, expressly rather than querying a generic legal database. The material the models work on is therefore the case file and the team's own work, so there is no case law database, statutory source, publisher or licensed reference set behind an answer.
One qualification is recorded because it points the other way and is unresolved: the vendor's own announcement states that the platform synthesizes case-level context including from e-discovery and legal research platforms, which implies material arriving from a research product without identifying it or addressing any license attaching to it. Recorded as the honest absence rather than a finding against the vendor. Searched the reachable home page material, the privacy policy and the December 2025 announcement on 5 September 2026.
Does the product tell you when the authority it just cited has been overruled?
No located public material addresses whether authority is checked for subsequent history.
Nothing that could be read addresses checking authority for subsequent history. Citations are described as verified inside the drafting surface, with no statement of what verification consists of. The architecture claim is that grounding comes from the matter's own record rather than a legal database, so primary law is not the product's own source. The vendor also states that the platform synthesizes context from legal research platforms and that its drafting surface produces filings.
Authority therefore does pass through the system, yet nothing describes a citator, a treatment signal or a currency check applied to it. That is a gap on a product whose output is meant to be filed. The home page material, the privacy policy and the vendor's December 2025 announcement were searched on 5 September 2026. The site blocks automated access, and its product pages were not opened and may say more.
What does the product do when the answer is not in the corpus?
No located public material addresses what the product does when it cannot ground an answer.
What the system does when it cannot ground an output is undescribed. There is no abstention path, no state in which it declines to answer, no confidence indicator and nothing on behavior where the case memory holds conflicting accounts. That is a routine condition in contested litigation, and the platform lists contested issues among what it tracks. The company's nearest statement is about sequencing rather than uncertainty: the system learns about the case before it does the work and is not permitted to skip that step.
That describes when the model acts, not how it behaves when the record is thin or inconsistent. Nothing describes what a user sees when an assertion cannot be supported from the matter. The site blocks automated access, tried on 5 September 2026, and its product documentation was not opened and may say more.
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 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.
The AI Hallucination Cases database maintained by Damien Charlotin was searched on 5 September 2026 on the product name Advocacy and on the domain advocacy.ai. No court order, opinion or disciplinary record naming the product or the company was located. One limitation applies to the search: advocacy is a common word in judicial writing, so a name-based search of this tracker is less discriminating for this vendor than for a distinctive product name, and a negative result carries correspondingly less weight. This records the state of the public record on that date and is not a finding about the product.
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
No located public material engages with bar or ethics guidance.
Professional conduct guidance gets no engagement at any level, from bar bodies or anyone else. No bar association, rule of professional conduct, ethics opinion or guidance for a particular jurisdiction is named or referred to in general terms. Nothing addresses a firm's obligations when a drafting surface produces material intended to be filed. Nor is there any nearby statement of the usual kind, since no advice disclaimer or statement that the company is not a law firm was located.
The founding team is drawn from practice, with a former securities litigator as chief executive and a former antitrust lawyer as chief operating officer, and the positioning rests on being built by litigators. The terms of service, where such material would usually sit, did not appear in search results, which is weak evidence of absence rather than proof. The site blocks automated access, tried on 5 September 2026, and its pages were not opened and may say more.
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Public materials claim time savings without addressing billing or disclosure, and the product sits inside a fee relationship between a lawyer and a client where those savings would change the bill.
Efficiency is claimed in the vendor's own framing and nothing addresses what it means for billing. The company describes the problem it solves as outcomes driven by bandwidth rather than talent. It states that when context compounds across a team, a firm is not just more efficient but operating at a strategic level, and it offers instant onboarding for new team members so that knowledge never walks out the door. Those are claims about time and leverage inside a firm.
No record of AI assisted work per matter is described, no guidance on fee or disclosure treatment is published, and nothing addresses what a client is told when case analysis and drafting came from accumulated machine memory. The buyers are Am Law and midsize firms billing hourly, so compressed associate time on research and drafting is where the disclosure question arises, and the material is silent on it.
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
No located public material supports a client side disclosure obligation.
Answering a client's AI clause would leave a firm with nothing to point to. The privacy policy names service providers by category only: a database provider for data storage, an analytics provider for website analysis and financial service providers for payment processing. It identifies no AI or model provider, on a platform whose value is entirely model driven. No subprocessor register, data processing addendum, consent pack or notification material was located, and no route to request any of it is offered.
The one document that would carry this material for a customer is the individual Customer Agreement the privacy policy defers to, which is negotiated per customer and unpublished. Nothing exists that a firm could forward to a client without first entering a commercial relationship. The site blocks automated access, tried on 5 September 2026, and any subprocessor list or DPA it publishes was not opened and may say more.
If a judge’s standing order requires an AI disclosure, can the product produce one?
No located public material addresses court disclosure or verification certification.
The product offers no way to produce a record of AI assisted work. One adjacent feature is worth noting. The platform keeps a visible activity history against docket entries, showing queries, reviews and notes with dates and the people responsible, and the company states that every query, review and note enriches the matter memory. That is a record of what the team did inside the workspace, not of what the model produced.
Nothing identifies a model or version against an output, nothing marks generated text once it reaches a draft, and no export is described for producing any of it to a court, a client or an opponent. No disclosure template or guidance was located. The gap matters because the drafting surface is promoted on the basis that what you draft is what you file, so the output is meant for a tribunal. The site blocks automated access, tried on 5 September 2026, and its product pages were not opened and may say more.
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.