Eve
AI platform built for plaintiff law firms across the full case lifecycle from intake to resolution, covering case intake and evaluation, medical overviews and chronologies, demand letters, complaints, propounding and responding to discovery, and case value analysis. Founded 2020 in San Francisco by Jayanth M, who serves as chief executive. In January 2026 the product was relaunched as an AI Workforce built around three coordinated agent roles, Agents, Auditor and Analyst, positioned to advance cases proactively rather than waiting for attorney prompts, and in June 2026 the company launched EveOS, described as an AI native operational platform for plaintiff firms. The platform is stated to reference all sources for its answers with links back to the firm's own documents, to proactively validate its responses, and to allow one click verification. It is also stated to learn a firm's tone, preferred language and style so that generated demand letters, complaints and discovery responses match the firm's voice, adapting further with use. Practice coverage is personal injury as the primary vertical, with medical malpractice, employment and general litigation also served. Security and privacy claims include SOC 2 Type 2 certification, described by the vendor as the first for a legal AI, HIPAA compliance, AES-256 end to end encryption, secure cloud infrastructure, and policies vetted annually by independent audits. The vendor states that the platform is built with attorney client privilege as a first principle, that case data is never used to train shared models, and that data is kept strictly isolated to the firm at organisation, user and workflow levels. Stated scale is more than 450 firms, more than 200,000 legal cases processed annually, and firms collectively recovering more than $3.5 billion in settlements and judgments. Funding totals $164m across three rounds: a $14m seed in October 2023, a $47m Series A in January 2025 led by Andreessen Horowitz, and a $103m Series B on 30 September 2025 led by Spark Capital at a valuation above $1 billion, with Lightspeed Venture Partners and Menlo Ventures participating. Headcount was reported at 314 as of June 2026. Pricing is not published.
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.
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.
AI native from founding and now architecturally explicit about it. The company was founded in 2020 to build this and has no preceding product to attach models to. The January 2026 relaunch as an AI Workforce is built around three coordinated agent roles, Agents, Auditor and Analyst, and the June 2026 EveOS launch positions the platform as an AI native operating layer for the firm rather than a tool inside one. Every deliverable is generated: intake evaluation, medical overviews, demand letters, complaints and discovery responses on both sides of the request. The system is also stated to adapt to a firm's writing style with use, which is model behaviour rather than configuration. Remove the models and nothing remains. Sixth A on this axis in the pull and the third consecutive one in this category, which now looks structural rather than coincidental: every plaintiff side vendor built so far was model first, because the underlying task of reading medical records at volume was never automatable any other way.
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.
Grounding is architectural, verification is built into the interface, and no measurement is published. The vendor states that all sources for Eve's answers are referenced with links back to the firm's own documents, that the system proactively validates its responses, and that a user can verify a response in one click. Proactive validation is a stronger claim than passive citation because it describes the system checking itself rather than only showing its work, and the Auditor agent role introduced in the January 2026 AI Workforce release suggests that checking is a named component rather than a background behaviour. Held at B because none of it is quantified or specified: no accuracy figure, no extraction recall on medical records, no hallucination rate, no evaluation, no description of what proactive validation actually tests or what happens when validation fails, and no statement of behaviour when the underlying records are incomplete or contradictory. A named auditor agent with no published output is a claim about architecture rather than evidence of accuracy.
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.
The most autonomous product in this category and the most explicit about it, with the oversight built as a role rather than asserted as a principle. The AI Workforce is described as three coordinated agents that advance cases proactively rather than waiting for attorney prompts, which is a real autonomy claim stated plainly rather than hedged, and EveOS extends that to an operational layer. The Auditor role is the notable design decision: a checking function is a named part of the workforce rather than a caveat, which is the correct architecture for autonomous case advancement. Held at B because the boundaries are entirely undescribed. Nothing states what an agent may do without attorney approval, whether anything is sent or filed unattended, what triggers Auditor review, what standard it applies, what happens when Agent and Auditor disagree, or how a firm configures the limits. Proactive advancement of a live case is the highest autonomy claim on this index and it arrives without a single published bound.
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.
The first A on this axis in the pull, and it is earned on the combination rather than on any single item. Scale is published and dated: more than 450 firms, more than 350 added in eight months, more than 200,000 legal cases processed annually, and firms collectively recovering more than $3.5 billion in settlements and judgments, all stated in a dated September 2025 funding announcement. Named customers with quantified outcomes and attributed comment: Frontier Law Center, with managing partner Manny Starr quoted by name in a named publication in March 2026, reporting intake conversion rising from 10 percent to 35 percent, intake process time falling by 50 minutes per call, and average case value increasing 90 percent. James Scott Farrin attorneys completing medical chronologies in minutes that previously took weeks. A separate verified user report of case capacity increasing up to 2.5 times without adding staff. An outsider can name the firm, name the partner, find the publication and check the date without contacting the vendor, which is the bar this axis sets. The figures still carry no methodology or baseline, and the $3.5 billion recovery total is a sum over customer outcomes rather than a demonstrated causal contribution, so the note records what the numbers are and are not.
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 FIRST RECORD IN 42 TO ADDRESS ATTORNEY CLIENT PRIVILEGE DIRECTLY. Every other vendor on this index, including the two plaintiff side competitors already built, graded C or B here on general security credentials with privilege simply unmentioned. Eve states that the platform is built with attorney client privilege as a first principle, that case data is never used to train shared models, and that data is kept strictly isolated to the firm at organisation, user and workflow levels. Three named isolation levels is a specification rather than an assurance, and workflow level isolation in particular addresses the case where one matter's content should not surface inside another. Paired with the credentials this category actually needs: SOC 2 Type 2 certification, HIPAA compliance for the claimant protected health information the product ingests, and AES-256 end to end encryption. Graded A because it is strictly more than the B records in this category, which carry the same class of credential without any privilege position at all. The isolation architecture is asserted rather than documented and no privilege specific attestation exists, which is why the note says so, but the ladder has to reward a vendor that answers the question over vendors that do not acknowledge it.
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.
Not located, and the gap is wider here than on any other record in this category because the autonomy claim is larger. Agents are stated to advance cases proactively without waiting for attorney prompts, and the product drafts complaints and discovery responses, which are filed documents bearing an attorney's signature and certification. Nothing published addresses the supervising attorney's duty over proactively advanced casework, what an attorney is certifying when signing a machine drafted pleading, the professional responsibility dimension of generated discovery responses, or any bar guidance. The Auditor agent is a quality mechanism inside the product and not a statement about the lawyer's own obligations. Checked the home page, the working with Eve page, the product material, the press coverage of both 2026 launches and the site navigation on 29 Aug 2026.
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.
Nothing published about how the models are governed, evaluated or monitored. No AI policy, no model card, no bias or fairness testing, no evaluation methodology or result, no accuracy monitoring, no drift statement, no named governance body and no ISO 42001. The category risk flagged on EvenUp and Supio applies here with an additional dimension: this product performs case evaluation and identifies missed case value, so it makes judgements about what a claim is worth, and it does so inside an agent architecture that advances cases proactively. A systematic tendency in valuation would propagate through 200,000 cases annually without an attorney necessarily prompting the judgement that produced it. The Auditor agent checks output within the system and is not an independent evaluation of the models. Checked the home page, the working with Eve page, the press coverage and the site navigation on 29 Aug 2026.
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.
A genuine stewardship position, disclosed with a qualifier the vendor did not have to admit. Stated: case data is encrypted with AES-256 end to end, never used to train shared models, and kept strictly isolated to the firm at organisation, user and workflow levels. Separately and openly stated: the system learns a firm's tone, preferred language and style and adapts further with use. Those two statements are consistent and together describe a specific architecture, being firm scoped adaptation with no cross customer pooling, and disclosing the second alongside the first is more candid than the vendors who state a training commitment and leave the learning behaviour unmentioned. Held at B rather than A because the specification stops short: no retention period is published for case content or generated output, no deletion right at matter close is described, nothing states whether firm scoped adaptation can be disabled, and nothing describes what happens to an adapted model when a firm terminates. HIPAA compliance covers protection of the claimant health data rather than its use.
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.
No published position located. Nothing was found on liability for AI output, warranty, service levels or remedy. The exposure is the largest in this category because the autonomy is: agents that advance cases proactively can act on a matter without an attorney initiating the step, and the product drafts pleadings and discovery responses subject to filing deadlines and certification requirements. A missed deadline, a defective discovery response or an understated case valuation all carry consequences that fall on the firm and ultimately on the injured claimant, and no published vendor position addresses any of them. Checked the home page, the working with Eve page, the press coverage and the site navigation on 29 Aug 2026. Enterprise agreements govern this and are not public.
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.
Nothing located, and this grade carries the weakest research basis on the record. No case management system integration, no document management connector, no records retrieval integration, no research provider and no API documentation were found across the home page, the working with Eve page, the product material, the press coverage of the AI Workforce and EveOS launches and the site navigation checked on 29 Aug 2026. The positioning offers one possible explanation and is not treated as an answer: EveOS is described as an AI native operational platform for the firm, which suggests the product intends to be the operating layer rather than a component that plugs into one, and a system positioned to replace the operational stack has less need to integrate with it. That is inference and is not credited. Recorded as a documented absence across the surfaces actually checked and flagged as the strongest correction candidate on this record. Compare Supio at B, which names Litify, MyCase and CasePeer.
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.
Nothing located beyond a generic phrase. Secure cloud infrastructure is stated and names no provider, no region and no residency commitment, and no single tenant or dedicated instance option is described. The firm level isolation the vendor does describe is a logical separation claim rather than a deployment or residency statement, and the two should not be conflated: isolated to your firm says nothing about which jurisdiction the data sits in. For a platform holding claimant protected health information at the scale of 200,000 cases annually, residency is a question a firm's own compliance review asks directly. Checked the home page, the working with Eve page, the press coverage and the site navigation on 29 Aug 2026. Compare Supio at B, which states data centres in three named countries.
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.
Named certifications with an annual cadence stated, and no route to anything underneath them. Published: SOC 2 Type 2 certification, HIPAA compliance, AES-256 end to end encryption, secure cloud infrastructure, and the statement that policies are vetted annually by independent audits, which addresses currency in a way most records in this category do not. Held at B rather than A on the familiar three absences: no auditing firm is named, no examination period, scope or certificate date is published, and no trust centre, security page or self serve documentation request route was located, so under the three tier test the artifact is absent rather than gated. One claim recorded and not credited: the vendor describes itself as the first legal AI to be SOC 2 Type 2 certified. That is a priority superlative rather than a security fact, it is not checkable from any published source, and it says nothing about the current scope or rigour of the examination. Calibration within this category: EvenUp reaches A on a Trust Center with a request route and a dated recertification; Eve and Supio both sit at B with claims and no route.
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.
Nothing located. No foundation model provider, model family or version is named, nothing states whether third party models sit underneath the agent architecture, and no subprocessor list was found. The commitment that case data is never used to train shared models establishes that a distinction between shared and firm scoped models exists in the architecture without identifying what either is built on. For a platform processing protected health information across 200,000 cases annually, the identity of every processor in the chain is a question a firm's HIPAA business associate diligence asks directly and it cannot be answered from public material. Checked the home page, the working with Eve page, the press coverage of both 2026 launches and the site navigation on 29 Aug 2026.
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.
No pricing published at any level. No price, no range, no unit of charge, and no indication of whether the model is per seat, per case, per matter or per firm, which matters because the product is positioned as an operational platform rather than a per document tool and the pricing unit would signal which. Independent material notes that the target segment is small to medium plaintiff firms without any published attorney count, revenue band or case volume threshold defining it, so a firm cannot self assess fit either. The contingency fee point recorded on EvenUp and Supio applies equally: case costs are advanced against a claimant's recovery, so undisclosed pricing is undisclosed cost to an injured person. Checked the home page, the pricing navigation, the working with Eve page and independent review material on 29 Aug 2026.
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.
Lifecycle coverage is the broadest in this category and the practice boundary is stated honestly. Coverage runs intake through resolution rather than centring on one document type: intake and case evaluation, medical overviews and chronologies, demand letters, complaints, and both propounding and responding to discovery, which is wider than either competitor built so far and includes the litigation phase rather than stopping at the demand. Practice areas named are personal injury as the primary vertical with medical malpractice, employment and general litigation, and independent review confirms no evidence of family, criminal, intellectual property or transactional work, so the scope claim holds. Held at B rather than A because the firm segment is asserted without definition, with independent material noting that the small to medium target is stated with no attorney count, revenue band or case volume threshold, and because nothing characterises depth: no record types or formats handled, no jurisdictional scope for the drafting, and no indication of whether discovery coverage extends across state and federal practice.
Legal Signals
What each signal meansA 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.
Client Data in Training
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.
Policy never for shared models, with firm scoped learning disclosed rather than hidden, which is the most fully specified position on this signal in the pull. READ THE QUALIFIER, then read what the vendor volunteers next to it. The quoted commitment is scoped to shared models, and the same page separately states that Eve learns a firm's tone, preferred language and style and adapts further with use. Those two statements are consistent and together describe firm scoped adaptation with no cross customer pooling, which is a real architecture rather than a hedge. What separates this from the LinkSquares pattern is candour: LinkSquares scoped its never to third party and public models and said nothing about its own, while Eve scopes its never to shared models and then openly describes the learning that happens inside the firm boundary. Isolation is specified at organisation, user and workflow levels. Held short of a stronger value because it is policy rather than contract, no opt out from firm scoped adaptation is described, no retention period is published, and nothing states what happens to an adapted model when a firm leaves.
Prompt and Output Retention
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.
Not addressed. No retention period is published for case files, medical records, generated demands, complaints or discovery responses, and nothing indicates whether a firm can require deletion at matter close. Retention has an extra dimension on this record because the vendor states the system adapts to a firm's writing style with use, which implies persistence of something derived from the firm's content beyond the individual matter, and nothing describes what that derived state is, how long it lives, or whether it can be reset. Checked the home page, the working with Eve page, the press coverage and the site navigation on 29 Aug 2026.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
Segregation is asserted in public materials with no published detail on how it is enforced.
Claimed and not documented, and it is the most specific claim of its kind in the pull. The vendor states that data is kept strictly isolated to the firm at organisation, user and workflow levels, which names three distinct boundaries rather than asserting isolation generically, and workflow level isolation is the one that speaks to matter separation directly. That specificity is why this records as claimed rather than as not addressed. What is missing is any documentation behind it: no description of how the boundaries are enforced, whether user level isolation means retrieval respects the asking user's permissions, whether a matter can be walled from other users within the same firm, or how the agent architecture interacts with those boundaries when agents advance cases proactively across a caseload. No document management system integration exists to inherit permissions from. Checked the home page, the working with Eve page and the press coverage on 29 Aug 2026.
Third Party Request and Subpoena Notice
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.
Not addressed. No government or law enforcement request clause, no commitment to notify a customer before producing their data, and no transparency report were located. The vendor holds claimant protected health information, unfiled pleadings and discovery work product across more than 450 firms and 200,000 cases annually, so a request served on the vendor rather than on a firm would reach material a firm would otherwise be positioned to resist producing. Checked the home page, the working with Eve page, the press coverage and the site navigation on 29 Aug 2026.
Primary Law Corpus Provenance
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.
Not addressed, and largely inapplicable in the usual sense with one live residue. The platform operates on the firm's own case files, medical records and documents rather than on a published law corpus, so there is no external legal source to name, license or date. The residue is what the underlying models were built on. The product drafts complaints and discovery responses, which requires familiarity with pleading conventions and procedural rules, and it performs case evaluation, which implies exposure to outcome data, and nothing published states what corpus supports either. Unlike EvenUp, no settlement or verdict repository is named as a distinct asset, so the basis for case value identification is unstated as well as unsourced. Checked the home page, the working with Eve page, the product material and the press coverage on 29 Aug 2026.
Good Law Verification
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.
Not addressed, and applicable rather than a scope fact. The product drafts complaints and discovery responses, which cite legal authority and are filed with courts, so a currency question genuinely arises and the stakes are higher than for the demand only products in this category since the output reaches a docket. Nothing published names a citator, a research provider or any treatment checking step, and unlike EvenUp and Supio no Thomson Reuters or Westlaw integration was located, so it is not clear from public material what source supports legal citation in generated pleadings at all. Third record in this category where this signal is live and the third to leave it unanswered. Checked the home page, the working with Eve page, the product material and the press coverage on 29 Aug 2026.
Refusal and Uncertainty Behaviour
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.
Not addressed, with one architectural component that gets close and stops short. The vendor states that Eve proactively validates its responses and enables one click verification, and the January 2026 AI Workforce release names an Auditor agent as one of three coordinated roles. Both describe checking rather than uncertainty behaviour: nothing states what happens when validation fails, whether the system declines to answer, flags low confidence, surfaces a gap in the medical record, or proceeds and marks the output. For an architecture where agents advance cases proactively without attorney prompting, the behaviour under uncertainty is the load bearing question, because an unprompted action taken on a weak inference is the failure mode a lawyer would never see coming. Checked the home page, the working with Eve page and the press coverage of both 2026 launches on 29 Aug 2026.
Fabricated Citation Record
Does a public court record exist involving output from this product?
No court order, opinion or disciplinary record naming this product has been located as of the date shown. This is a statement about the public record, not a finding about the product.
None located, with the instrument named. General web searches combining the vendor and product names with court, order, sanction, fabricated citation and pleading terms returned nothing on 29 Aug 2026, and no named docket database or court record tracker was searched. Recorded as a statement about what this search found and not as a clearance. The exposure shape is the broadest in this category: this product drafts complaints and discovery responses that are filed with courts under an attorney's signature rather than demand letters sent to insurers, so a defective citation would reach a docket directly and become part of the public record. Flagged as the strongest candidate in this category for a proper docket search on a later pass.
Bar Guidance Alignment
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.
Not addressed. No named ethics opinion, no ABA Formal Opinion 512, no state bar guidance and no engagement with professional conduct rules was located. Relevance is at its highest in this category: the product drafts pleadings and discovery responses filed under an attorney's certification and operates an agent architecture that advances cases without prompting, which engages competence, supervision and candour duties simultaneously. Checked the home page, the working with Eve page, the product material, the press coverage and the site navigation on 29 Aug 2026.
Billing and Fee Posture
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.
Savings claims only, and the published figures reach further than throughput. Reported: intake conversion rising from 10 percent to 35 percent and intake process time falling by 50 minutes per call at a named firm, case capacity increasing up to 2.5 times without added staff, and average case value increasing 90 percent. The last of those is a claim about client recovery rather than firm efficiency, which is unusual and consequential, and it carries no methodology, baseline or comparison group. Nothing appears on the client's side of the equation: no position on whether platform cost is a case expense or firm overhead in contingency work, no disclosure guidance, and no record a firm could produce showing what portion of a demand, complaint or discovery response was machine generated. Checked the home page, the working with Eve page and independent review material on 29 Aug 2026.
Outside Counsel Guideline Readiness
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.
Not addressed. Certifications are claimed, being SOC 2 Type 2 and HIPAA, alongside AES-256 encryption and annually audited policies, which gives a firm citable content for a client questionnaire. No route to anything underneath was located: no trust centre, no security page, no request path for the SOC 2 report, no subprocessor list, no named model provider and no business associate agreement material. A firm could repeat the claims and could not evidence them or answer which third parties process claimant health data. Recorded as not addressed on the same basis as Supio, and the distinction from EvenUp at on request remains the existence of a request route rather than the strength of the claims. Checked the home page, the working with Eve page and the site navigation on 29 Aug 2026.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
Some elements of the record are available, short of a document level export.
Partial record. The source trail is real: all sources for Eve's answers are stated to be referenced with links back to the firm's own documents, with one click verification, so a party can show what a factual assertion rests on and open the underlying record. The two familiar limbs are missing and one of them matters more here than anywhere else in this category. Nothing indicates that output records which model generated it, and no human verification record is captured, so a firm cannot evidence that an attorney reviewed a machine drafted pleading before filing it. That is the exact question a standing order on AI use in court filings asks, and this is the only product in the category that drafts complaints and discovery responses reaching a docket rather than demand letters reaching an insurer. The Auditor agent performs checking inside the system and no artifact of that check is described as reaching the customer or the court.