C
Centari

Centari is a deal intelligence platform for transactional law firms and investment funds. Its Deal Reasoning Engine converts deal documents such as credit agreements, limited partnership agreements and M&A closing sets into a structured, citation-backed database built to each firm's own taxonomy, which lawyers use to find precedent clauses, benchmark negotiated terms, profile matters and prepare pitch materials. Features added through 2026 include Projects for one-off extraction grids, a Centari Intelligence sidebar that answers questions with citations to source language, Tasks for tracking human verification of extracted data, External Views for sharing white-labelled dashboards with clients, and an MCP server with a Microsoft Copilot integration.

A dedicated account team of attorneys and engineers calibrates each deployment. Centari, Inc. is a venture-backed Delaware corporation headquartered in New York, and the firms it names as customers include Fried Frank, Willkie Farr & Gallagher, Ropes & Gray, Wilson Sonsini and Gilbert + Tobin.

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.

AA on AI CentralityThe artificial intelligence is the product. Remove the models and there is nothing left to sell.

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.

Centari sells the database its AI builds. The Deal Reasoning Engine reads a firm's credit agreements, limited partnership agreements and M&A closing sets and turns them into structured, citation-backed data points, and the platform overview describes multi-stage reasoning that resolves defined terms, traces cross-references and evaluates conditional logic in sequence rather than skimming a document as flat text. Everything else a firm buys runs on that output: precedent search, market benchmarking, Views dashboards, Projects grids and the Centari Intelligence sidebar.

The MCP server and Microsoft Copilot integration let a firm's other AI tools query the same data, and they belong to the integration picture rather than to what Centari itself generates. Remove the models and there is nothing in the database to search or chart. Verified 20 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.

Every extracted value comes back with a citation, and nothing about accuracy is measured. The platform overview says each data point traces to its source language and connects to related provisions across the deal documents, with a verification view for checking it, and the 24 August 2026 release post says the Centari Intelligence sidebar answers with citations to the source language that a partner can open in one click.

The method behind the extraction is described rather than asserted: defined terms resolved, cross-references traced, conditional logic evaluated in sequence. What a buyer cannot weigh is performance. The site claims attorney-level precision and unmatched accuracy with no test set, no score and no failure mode named, so the only check on a given field is the firm's own reviewer. The citations run to the firm's own agreements rather than to primary law. Verified 20 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.

Centari is built around people checking what the AI extracts, and it gives a firm a way to track that checking. The 24 August 2026 release post describes the product as designed for human in the loop verification and introduces Tasks, which gathers every open verification step across Databases and Projects so an owner can assign reviewers and see which data points a person has checked. The platform overview describes a verification view beside each extraction.

Projects are kept apart from the governed record: nothing in a Project writes back to a Database. What is not published is where the system itself draws a line. Nothing says when extraction stops or flags low confidence, and nothing says what unverified output must not be used for, so those rules are left to each firm. Verified 20 September 2026.

Source: Vendor Published
BB on Operational and Outcome EvidenceReal deployment evidence with substance, short of full attribution or measurement: a named customer without figures, or figures without the named customer.

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.

Centari names large law firm customers and puts senior people on the record, but publishes no measured result. The homepage, platform overview and For Law page carry attributed endorsements from Fried Frank (Global Head of Innovation and Knowledge), Willkie (Counsel, Asset Management), Ropes & Gray (Counsel), Wilson Sonsini (Senior Manager of KM) and Gilbert + Tobin (Partner, Corporate Advisory), and from Sentinel Global's general counsel on the fund side.

The blog lists a Gilbert + Tobin case study dated 30 April 2026 describing a firm wide M&A deployment, and announcements of the Fried Frank and Ropes & Gray adoptions. No figure for time saved, accuracy or outcomes is published. The case study links on the site did not open a readable page on 20 September 2026, so the detail of the Gilbert + Tobin deployment could not be read. What is on the record is endorsement by named senior users, not a measured result. Verified 20 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.

The commitments about client material are specific, and they live on a security page rather than in a contract. The security page states that customer data is siloed and never commingled, with single tenant deployments available, that it is never used to train or fine tune the underlying models, that it is encrypted in transit and at rest with bring your own key supported, and that access can be controlled down to the data point with SSO and ethical wall integration.

The Terms of Use dated 16 September 2025 say nothing about customer content and state that they sit alongside any other agreement; that other agreement is not published, so a buyer cannot read any of these promises as a term before signing. Nothing addresses privilege or work product in the deal documents a firm uploads, how long they are held, or what OpenAI and Anthropic, named as AI providers in the Trust Center, may retain.

For a product that ingests a firm's whole deal history, that last gap is the one a general counsel will raise first. Verified 20 September 2026.

Source: Vendor Published
BB on UPL and Professional Responsibility PostureA real position is published on advice versus tooling, short of full treatment: commonly a disclaimer without the supervision and competence dimension, or silence on jurisdiction limits.

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. 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.

Section 4 of the Terms of Use, dated 16 September 2025, is direct about what Centari is not. It states that the company is not a law firm, attorney, staffing firm or lawyer referral service, that nothing on the service is legal advice, that no attorney-client relationship arises, and that employees who are licensed attorneys are not practising law in that capacity. The audience is professional throughout: transactional practices, knowledge teams, in-house legal and investment funds.

The other half of the question is untouched. Nothing addresses how a supervising lawyer should treat extracted terms that feed advice, negotiation positions or the External Views dashboards a firm shares with its own clients, and no jurisdiction limits are stated. Verified 20 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.

What Centari publishes is how a firm checks the AI's work, not how Centari builds and controls it. Calibration of each firm's taxonomy by domain experts is described on the homepage and the For Law page, an account team of attorneys and engineers runs it, and the 24 August 2026 release post describes the Tasks verification workflow. That tooling helps a firm govern its own use; it says nothing about how Centari tests and controls its models.

No accountable owner inside the company is named, nothing describes what is tested before a release ships, no evaluation result is published, and nothing addresses whether extraction quality is uneven across document types, drafting styles, jurisdictions or languages. Checked the homepage, platform overview, For Law, security and about pages, the blog and the Trust Center on 20 September 2026. Verified 20 September 2026.

Source: Vendor Published
CC on AI Safety and Data StewardshipA generic privacy policy covers the product without addressing what happens to documents and prompts after processing.

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.

The controls protecting the data are published; what happens to a firm's documents afterwards is not. The security page states encryption in transit and at rest with bring your own key, siloed data with single tenant options, access control down to the data point, SSO and regular penetration testing, and the Trust Center lists security controls including a record of incidents and a subprocessor list. No retention period for uploaded deal documents or for questions put to the product is published, no deletion commitment applies on termination, and no incident notification timeline was located.

The Privacy Policy dated 16 September 2025 covers personal information collected through the website and services rather than the contracts a firm loads, and no customer agreement or data processing addendum is published. Verified 20 September 2026.

Source: Vendor Published
CC on AI Liability and RecourseLiability is addressed only through a standard limitation clause that disclaims the exposure the product creates.

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.

What Centari refuses is published in full; what it stands behind is not. The Terms of Use dated 16 September 2025 provide the service as is and disclaim warranties of accuracy, reliability and fitness (section 15), cap total liability at the greater of fees paid in the prior six months or one hundred US dollars (section 16), and require the customer to indemnify Centari (section 17). No indemnity runs the other way, no warranty attaches to extracted output, and no insurance position appears.

The Terms state that they sit alongside any other agreement, and that agreement, where an enterprise buyer's real allocation of loss would sit, is not published, so the only readable position is the one that allocates the risk of a wrong extraction to the firm. Verified 20 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.

The connections are named and none of them is documented. The platform overview says Centari's MCP Server makes a firm's deal intelligence available to the AI tools it already uses, the blog lists a 16 June 2026 announcement of that server and a one-click Microsoft Copilot integration, and the security page names SSO and ethical wall integration. There is no integrations page, no developer documentation and no API reference, and the June announcement does not open a readable page.

Nothing states how deal documents get from a document management system such as iManage or NetDocuments into Centari, which is the step a firm has to plan and staff, or what the Copilot integration exposes to a user's assistant. Projects grids are described as exportable. Verified 20 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.

The tenancy model is stated and the geography is not. The security page says customer data is siloed and never commingled, that single tenant deployments are available, that bring your own key encryption is supported, and that regional hosting and processing are available for customers around the world. No region is named, nothing says which arrangement or tier carries single tenancy, and nothing separates where the database sits from where OpenAI and Anthropic process the text of a firm's agreements.

A firm with a UK or EU residency requirement would have to establish all of that in a sales conversation. Verified 20 September 2026.

Source: Vendor Published
BB on Security Certifications and Trust CenterCertification is real and stated, short of accessible evidence: a named standard without scope, date, or a way to obtain the report.

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.

Centari states SOC 2 Type II and ISO 27001 certification and regular third party penetration testing on its homepage and security page, and links a Trust Center hosted by Sprinto. Only part of the Trust Center could be read on 20 September 2026. The part that could be read shows a compliance section with two entries marked compliant, a list of controls and a subprocessor list. It names no auditor, report period or scope, and no route to obtain the reports, so the certifications stand as the vendor's statement until a buyer sees the reports. Verified 20 September 2026.

Source: Vendor Published
BB on Model Supply Chain DisclosureThe supply chain is partly disclosed: providers named without change notification, or architecture described without the providers.

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 providers are named; the models are not. Centari's Trust Center carries a subprocessor list naming OpenAI and Anthropic as AI providers and Google as a cloud provider. Only part of that list could be read on 20 September 2026, so it may name others. The marketing pages refer to 'our underlying AI models' and to a patent pending Deal Reasoning Engine without identifying either: no model, no version, and nothing about where the AI providers run.

No commitment to tell customers when the models behind their extractions change was located on the security page, the platform overview or the Trust Center, which matters here because a firm's benchmarks are built on extractions made by whatever model was running at the time. Verified 20 September 2026.

Source: Vendor Published
DD on Commercial TransparencyNo pricing information published at any level, including the unit of charge.

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.

Nothing about price is published, including the unit of charge. Checked the homepage, the full navigation and footer, the platform overview, the For Law and For Finance pages, the security and about pages, the Terms of Use and the blog on 20 September 2026: there is no pricing page, no tier names, no unit, no trial and no figure anywhere. Every route offered is a demo request. Section 13 of the Terms of Use reserves the right to charge fees for the services without stating what they are. Verified 20 September 2026.

Source: Operator Verified
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.

Who the product is built for is set out clearly; where it stops is not. The For Law page addresses transactional practices and names partners, knowledge management, associates and business development as its users, and the For Finance page addresses investment funds across legal, investor relations, finance and compliance. Document types are named rather than implied: credit agreements, limited partnership agreements and M&A closing sets, with the work described as high stakes M&A, fund formation and financings.

The demo form offers law firm, in-house legal, financial services and government. Nothing states what falls outside, such as litigation material, agreements in other languages, or deal types for which the firm's taxonomy has not been calibrated, and calibration is how every deployment starts. Verified 20 September 2026.

Source: Vendor Published
Sources on file

6 public documents

The public pages on file for Centari, 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.

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.

The security page and homepage state that customer data is never used to train or fine-tune Centari's models. No published agreement carries the commitment: the Terms of Use (16 September 2025) are silent on customer content and supplement an unpublished customer agreement.

Source: Vendor PublishedYour data is never used to train or fine-tune our underlying AI models.As of Sep 20, 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.

Checked the security page, the platform overview, the Terms of Use, the Privacy Policy, the 24 August 2026 release post and the Trust Center on 20 September 2026. No retention period is stated for uploaded deal documents, for questions put to the Intelligence sidebar or for extracted output, and no deletion commitment applies on termination; the security page says only that a customer's data is always under its control.

Source: Operator VerifiedAs of Sep 20, 2026

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.

The security page claims access control down to the data point, SSO and ethical wall integration, and siloed customer data. No published material describes which systems the wall integration reads from or how access is enforced at query time.

Source: Vendor PublishedCentari supports SSO and ethical wall integration.As of Sep 20, 2026Evidence

Third Party Request and Subpoena Notice

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

Disclosure addressed, notice absent

Published terms or policy address disclosure to authorities or in response to legal process, and no commitment or reservation regarding customer notice is located anywhere. The vendor has told the customer that data can leave and has said nothing about whether the customer hears of it.

The Privacy Policy (16 September 2025) permits disclosure of information to comply with legal process and says nothing about notifying the customer. It covers personal information; no published term addresses requests for customer deal documents, and the Terms of Use are silent on the point.

Source: Vendor Publishedto comply with legal processAs of Sep 20, 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?

Not addressed

No located public material identifies the corpus behind the product’s answers.

Checked the homepage, the platform overview, the For Law page and the security page on 20 September 2026. Centari works over the deal documents a firm loads and its benchmarks are drawn from that material; no external legal corpus is described, so provenance here is the firm's own.

Source: Operator VerifiedAs of Sep 20, 2026

Good Law Verification

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

Not addressed

No located public material addresses whether authority is checked for subsequent history.

Checked the homepage, the platform overview, the For Law page and the 24 August 2026 release post on 20 September 2026. Centari cites clauses in the firm's own agreements rather than legal authority, so the question of subsequent history does not arise for this product and nothing addresses it.

Source: Operator VerifiedAs of Sep 20, 2026

Refusal and Uncertainty Behaviour

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

Not addressed

No located public material addresses what the product does when it cannot ground an answer.

Checked the homepage, the platform overview, the For Law page, the security page, the 24 August 2026 release post and the Trust Center on 20 September 2026. Nothing describes what happens when the engine cannot find a data point or the Intelligence sidebar cannot answer, and no confidence or grounding score is exposed. Verification runs through Tasks, which routes finished output to a human reviewer rather than describing any point at which the system declines to answer.

Source: Operator VerifiedAs of Sep 20, 2026

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 the AI Hallucination Cases database maintained by Damien Charlotin on 20 September 2026 on the product name Centari and the corporate name Centari, Inc. No court order, opinion or disciplinary record naming the product was located. This is a statement about the public record rather than a finding about the product, and the exposure is structurally lower than for a research tool: Centari cites a firm's own agreements and is sold for transactional work rather than for court filings.

Source: Bar Guidance or Court RecordAs of Sep 20, 2026Evidence
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.

Checked the homepage, the platform overview, the For Law page, the security and about pages, the Terms of Use and the blog on 20 September 2026. Section 4 of the Terms states that Centari is not a law firm and gives no legal advice, but no material engages with ABA Formal Opinion 512, a state bar opinion or any other named ethics guidance on the use of AI.

Source: Operator VerifiedAs of Sep 20, 2026

Billing and Fee Posture

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

Savings claims only

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.

The For Law page says associates find on-point precedent in seconds to bring to drafts and redlines, and the platform page contrasts the engine with manual work, which is work billed to clients on live deals. Much of the product also serves pitch and knowledge work that bills no client. No material addresses fee or disclosure treatment.

Source: Vendor PublishedAs of Sep 20, 2026Evidence

Outside Counsel Guideline Readiness

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

Subprocessors listed

A current subprocessor or model provider list is published.

Centari's Trust Center carries a subprocessor list that names OpenAI and Anthropic as AI providers and Google as a cloud provider. Only part of the list could be read on 20 September 2026, so it may name others. No material written for a firm to pass to its own clients, such as an AI use summary or a response to outside counsel guidelines, was located.

Source: Vendor PublishedAs of Sep 20, 2026Evidence

Court Disclosure Support

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

Partial record

Some elements of the record are available, short of a document level export.

Every extracted data point cites its source language, and Tasks records which data has been verified by a human reviewer. The product serves transactional work and nothing addresses court disclosure or records which model produced an output.

Source: Vendor PublishedAs of Sep 20, 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 303 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 20, 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