The Brief· September 5
The week in one line

Harvey named the model behind its product and published the retention and residency caveats with it, Wordsmith and Everlaw wired legal AI into the systems that hold the documents, and Avvoka started building templates from what a firm actually signed rather than what a partner remembered. The first issue of this log is mostly about where the work already lives. Read the issue →

Independent Reference · Legal Technology

The index of AI vendors in legal

The AI Legal Index is an independent index that tracks changes to AI vendors in legal. It grades the vendor and its product rather than the underlying model, the law firm, or the lawyer, which is what separates it from the other rankings in this territory. Legal is the one professional market where a confident wrong answer is a disciplinary matter, so every vendor is assessed on the same fifteen capability axes and twelve legal signals, built for the questions a general counsel, a firm risk committee and a practice group leader ask before they sign: what happens to privileged material, where the law in this product came from, what it does when it does not know, and whether the vendor can produce the record a judge might ask for. Everything is drawn from public sources, every record carries a verification date, and no vendor pays for inclusion, placement, or rating.

Change Wire

What changed at these vendors, this week

Full change log →

Every entry is dated, attributed to a named vendor and carries a live source. Product announcements that change nothing a buyer can verify are not logged, and neither are funding rounds, valuations or awards.

Sep 10, 2026
Intapp released a DealCloud with Celeste plug in for ChatGPT for Financial Services, available to eligible Intapp clients from 10 September. The plug in lets a professional query the firm's own client, deal, investor and relationship records from inside ChatGPT, with Intapp's information barriers, material non public information rules and independence obligations enforced on every answer, so a user sees only what they could already see in DealCloud. This first release is scoped to financial services firms on ChatGPT for Financial Services; Intapp has not said when, or whether, the same plug in reaches DealCloud for Legal.
Medium
Sep 10, 2026
Harvey introduced Contract Review Agents, in early access through a waitlist. A team builds an agent from its executed contracts, negotiation guidelines and playbooks, then answers a guided interview about how it applies those standards in practice, such as the liability cap it proposes and the fallback it will accept, or which clauses always escalate; the agent then delivers a first pass review of each redline with recommended edits, reasoning and citations to the sources behind each decision. Precedent benchmarking compares a counterparty's proposal clause by clause against positions the team has accepted in executed contracts, and the agent flags where recent negotiated outcomes have drifted from its own standards and proposes an updated position for the team to review.
Medium
Sep 9, 2026
Harvey's 9 September release note puts user level Memory into the product across the Assistant, the Word add in and Playbooks, two weeks after its 26 August early access announcement. A memory is a standing instruction such as a drafting style, a negotiating posture or a jurisdiction of admission, saved on request or when a prompt signals a preference, confirmed on a card at the point of saving, and then applied in future sessions across the web app, Outlook and Word, with Harvey citing the memory it used the way it cites a source. Every memory sits in Personal Settings where it can be edited, searched, deleted or switched off, admins control whether Memory is on and for whom, and Harvey states that individual memories are never used to train models.
High
Sep 8, 2026
Harvey released Agentic Vault Search to general availability and opened Agentic Vault Organization in Early Access. Search lets a user attach a whole Vault, or work inside a Space with Vaults attached, and have the Harvey Agent locate the relevant documents from signals such as folder placement, naming convention and uploader rather than exact keywords, operating under the user's own Vault permissions and ethical walls with every file it touches logged to the workspace audit trail. Organization takes a plain language instruction such as sorting a data room by counterparty and agreement type, proposes a plan showing every rename, move and tag before anything changes, executes only on confirmation, can undo the most recent set of changes, and never edits file content.
Medium
Sep 8, 2026
Lexroom acquired Query Juriste, a French legal AI research company, and Praven Intelekt, a Bulgarian legal AI assistant, in its first acquisitions since founding. Query Juriste brings a research platform connected to more than five million French and European legal documents, with roughly 300 customers and 1,000 paying users; Praven Intelekt brings more than 900 customers and Lexroom's first presence in Central and Eastern Europe. Both founding teams join Lexroom, the acquired brands transition to Lexroom over time, and the company now operates in Italy, Germany, Spain, France and Bulgaria.
Medium
Sep 4, 2026
LegalOn published a set of workflow additions: Cross Document Review for analysing a whole contract family together, searchable negotiation history through the LegalOn Assistant, and an upgrade to AI Revise that drafts and inserts language for missing provisions inside the Word add in. Teams can also set a department level editor name for redlines in Word for the Web.
Medium
What this index grades that others do not

Two axes nobody else in this territory grades

Thirteen of the fifteen axes are the standard the family applies to every vendor market it covers. Two exist only here, because they are the two questions that decide whether a legal AI product can be used at all rather than how well it performs.

AI Capability

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.

Compliance and Risk

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 Legal Signals

Twelve things recorded on every vendor, not graded

Every signal in full →

Some questions do not have a grade, they have an answer. Whether a vendor trains on client data is not a B. Whether a court record exists is a fact about the public record. These are recorded with a source and a date, and the value set for each one is published so a reader can see exactly what the alternatives were.

Confidentiality and Privilege

What happens to client material once it enters the product, and what the vendor has committed to in writing.

  • Client Data in Training
    Can material a lawyer puts into this product be used to train a model?
  • Prompt and Output Retention
    How long does the product keep what a lawyer typed, and can that be set to zero?
  • Ethical Walls and Matter Segregation
    Does retrieval respect the firm’s ethical walls, or can the model read across them?
  • Third Party Request and Subpoena Notice
    If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?

Accuracy and Authority

Where the law comes from, whether it is still good law, and what the product does when it does not know.

  • Primary Law Corpus Provenance
    Where does the law in this product come from, and does the vendor have the right to use it?
  • Good Law Verification
    Does the product tell you when the authority it just cited has been overruled?
  • Refusal and Uncertainty Behaviour
    What does the product do when the answer is not in the corpus?
  • Fabricated Citation Record
    Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?

Professional Responsibility

The obligations that attach to a lawyer rather than to software, and whether the vendor has engaged with them in public.

  • Bar Guidance Alignment
    Has the vendor engaged in public with the ethics opinions its buyers are bound by?
  • Billing and Fee Posture
    Does the vendor address what happens to the bill when the work takes an hour instead of six?
  • Outside Counsel Guideline Readiness
    Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
  • Court Disclosure Support
    If a judge’s standing order requires an AI disclosure, can the product produce one?
Standing

What a grade on this index means, and what it does not

A grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check, and the index says so rather than crediting a claim it cannot test.

The reverse is also true and worth stating plainly. A vendor with no located court record is not thereby safe, and a vendor that publishes its accuracy figures is not thereby accurate. The index measures disclosure and evidence, which is the only thing an outside party can honestly measure, and it dates every judgement so a reader can see how old it is.

Sourcing

Every record carries one of five source bases, shown on the record itself.

  • Vendor Published. The vendor’s own documentation, terms or trust materials.
  • Third Party Estimated. An outside source the index could corroborate, marked as an estimate.
  • Practitioner Interview. A named practitioner speaking to use in production.
  • Bar Guidance or Court Record. An ethics opinion, a court order, or a disciplinary record, cited.
  • Operator Verified. Checked directly in the product or against a live artifact.

9 categories, 15 axes and 12 signals on every published vendor.

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 61 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 15, 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