W

Wexler AI

The models are the whole product. Fact extraction across a document set, contradiction detection, chronology building and the Kim assistant all come from the models. The platform is positioned as reasoning across a million pages, not as a repository with analysis added on top. No document management, workflow or matter system sits underneath that would work without them. The company calls it a fact intelligence platform that analyzes every fact in every document to show what the evidence supports, and none of that could run without the models.

Grounding is documented and central to the product's claim. Answers are presented as sourced facts drawn only from the customer's own documents, and every fact traces to its source passage. An arbitration case study records the system explaining where in a document it found each date or event. What is missing is a measurement published by the vendor. The one accuracy figure on the site comes from an unnamed national firm, quoted in a testimonial.

In its own testing the hallucination rate was less than 0.1 percent, and the tool was at least on a par with human review. No test set, method or named organization stands behind that figure. The vendor names no failure mode anywhere.

Oversight is never described as a structure, though the design gives a reviewer tools to check the output. Confidence scores and source links on each answer give a reviewer something to test. A testimonial describes the system raising follow up questions of the kind a reviewer would expect from a trainee. Nothing published says what the system does unattended and what a lawyer approves. No threshold is given at which it defers or stops, no review workflow is described, and nothing commits to a supervising lawyer working alongside it. With no terms published, there is no contractual review duty either.

Deployment evidence is real, and the figures attach to unnamed firms. Four firms appear as named customers, among them Clifford Chance, Goodwin and Burges Salmon, with attributed quotes from Claire Freeman, a partner at Clifford Chance, and Tom Whittaker, a legal director at Burges Salmon. Lord Wilson of Culworth also endorses the product. The insights journal dates two relationships: Goodwin deepened its partnership on 3 June 2026, and Burges Salmon extended its partnership across the firm on 11 May 2026.

The figures are specific. One AmLaw 100 matter covered more than 250,000 pages with a 90 percent average time saving, and 100 percent of users said they would adopt the tool. At a global firm, inconsistencies the system identified were relied on at an international arbitration hearing. No figure is tied to a named firm.

Confidentiality is asserted in general terms, and no commitment can be read in advance. The security page speaks of a deep commitment to best in class security and a heavy emphasis on data protection, secure operations and cloud security, which is assertion rather than commitment. Nothing published addresses training on customer documents, retention or deletion of matter material, separation between customers or matters, or what a model provider may retain.

Privilege and work product go unmentioned, on a product built to take in the entire evidential record of a dispute. The privacy policy does not reach customer documents, and no agreement is published.

The line between analysis and advice is not drawn anywhere, for output relied on in filings and at hearings. No statement says what the output is and is not, no disclaimer separates analysis from advice, and no competence or supervision language appears. No jurisdiction limit is stated, though the product is sold in the United Kingdom and the United States. The gap is not an unread document, since no terms of service, customer agreement or acceptable use policy exists on the site. The marketing points the other way, describing the output as testable at trial.

No governance position is published for a system whose output shapes what a party argues the evidence shows. There is no responsible AI statement, governance framework or named owner. Nothing describes testing or evaluation before release, no results are published, and nothing addresses bias or uneven performance across document types, languages or kinds of matter. The confidence score is a product feature rather than a governance mechanism, and the information security certifications do not reach these questions.

A privacy policy for the website covers the company, not the evidence customers upload. It names contact details, IP address, browser and device data and usage data. None of it reaches uploaded evidence, extracted facts or the assistant's outputs. Matter material gets no deletion or retention terms, and no breach notification practice appears. The one specific stewardship practice published is a responsible disclosure program: security reports get a response within two business days, good faith researchers get a safe harbor, and credit is given on resolution.

Nothing is published on who bears the loss when the system is wrong. The site carries no terms of service, master subscription agreement, customer agreement or order form template. Its only legal link, in the footer, is a website privacy policy, and the navigation has no legal section. No indemnity, liability cap, warranty or insurance position appears on any page, so the allocation of loss cannot be read before signing. The home page, security page, privacy policy, navigation and footer were searched.

The pages checked show no integration with practice systems. No document management system, eDiscovery platform, review tool, transcript system or matter management product is named, and neither the navigation nor the footer leads to an integrations page or developer documentation. Single sign on, listed with the security features, is authentication rather than integration. The workflow described is to upload a document set and question it, with nothing on how documents arrive from an existing review platform.

Residency is offered as a real choice, and processing is not addressed. The home page and security page both say data is stored in the UK, the US or elsewhere on request, so the location is the customer's choice rather than fixed. Private cloud deployment is offered for organizations that need it, a tenancy option beyond the default. Nothing says where processing or model inference happens as distinct from storage. No regions are listed beyond the two named, and nothing explains how controls, isolation or price differ between the standard and private cloud arrangements.

Two certifications are claimed, and the evidence behind them is not public. ISO 27001 certification and SOC 2 Type II accreditation appear on the home page and again on a security page, alongside single sign on and regular penetration testing. No certifying body, auditor, certificate number, scope, observation period or report date is published, so neither claim can be checked against a register without contacting Wexler.

A trust center at trust.wexler.ai, hosted on Drata, is linked from the security page and described as supporting all customers. It showed only a loading frame with no content when checked, so what it holds and who can open it are not established. Current as of 4 September 2026.

Which models process a customer's documents is not disclosed. No model is named, no model provider is named, and no subprocessor list exists. No location is given for inference as distinct from storage, and nothing commits to notifying customers of a change. The assistant carries a product name, Kim, which identifies the interface rather than what sits underneath it. The security page mentions cloud security, and the residency statement says where data rests, but neither says whose models touch it.

No pricing is published at any level, including the unit of charge. There is no pricing page in the navigation or footer, no tier names and no feature split. Nothing says whether charging is per user, per matter, per page or per document, and there is no free trial or self serve route. Every call to action leads to a demo booking or a contact form asking for name, email, company, title and team size.

Segment and practice coverage are described with substance, and the edges are left open. The buyer is consistently disputes and investigations teams working on complex disputes, and the named customers and case studies are law firms, including one in the AmLaw 100. The roles quoted show the product used for early case assessment, international arbitration and cyber work. No firm size floor or ceiling is stated, and in house and government use are not addressed.

No jurisdiction of substantive coverage is named despite offices in London and New York, and nothing says which matter or document types the product does not handle.