C
Casefleet
Casefleet is case chronology and fact management software for litigators, operated by LAWPRCT, Inc. of Atlanta, Georgia, and sold to plaintiff and defence firms since 2016. The core of the product is a fact chronology in which every fact is linked to the source document it came from, supported by a document reviewer for highlighting and extracting facts as you read, visual timeline building, case outlines for structuring issues and arguments, full-text search across every page of every document, deposition transcript markup, audio and video review with transcription, and litigation-ready reporting.
The AI layer is branded Casey and ships as several named features: a case-aware assistant that answers questions in natural language and returns clickable citations linking back to source documents; Proposed Facts, in which Casey drafts facts from documents with a citation to each source and the user reviews, edits and approves every one before it joins the chronology; Document Intelligence covering summarisation, named entity recognition and semantic indexing; bulk AI tagging and renaming for document organisation; AI-enhanced search; and agentic workflows that hand Casey multi-step case tasks.
Practice-area coverage includes employment, criminal, personal injury, workers' compensation, family law, complex litigation and medical chronologies, alongside workflow pages for early case assessment, summary judgment preparation, cross-examination preparation and investigations, and a set of non-litigation uses aimed at individuals handling their own matters. Pricing is published in full, with a Starter plan at 30 dollars per user per month, an Advanced AI plan at 140 dollars, custom Enterprise quotes above twenty users, and per-unit overage rates published for optical character recognition, transcription, document intelligence, storage and AI credits.
The customer agreement, effective February 2026, carries a defined term for AI Features, a dedicated section of AI limitations, and a commitment that customer data will not be used to train Casefleet's AI systems without explicit consent.
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.
The AI layer is substantial and prominently positioned, and the vendor's own pricing table proves it is severable. The homepage runs under the label AI case management software and the features navigation opens with an AI-Powered group of seven entries, each with its own page: Casey AI Assistant, Agentic Workflows, Document Intelligence, Summarization, Entity Extraction, Document Organization and Proposed Facts. Against that, the published plan comparison shows the Starter plan at 30 dollars carrying none of it: document summarisation, semantic indexing, named entity recognition, fact extraction, natural language search, AI issue generation and AI credits are all marked unavailable on Starter and appear only on Advanced AI at 140 dollars.
A buyer can therefore purchase a fully working version of this product with no machine learning in it, which is the clearest severability evidence any record in this pull has produced. What remains on Starter is the 2016-era platform that the company built its name on: chronologies, timelines, outlines, document review, transcripts and reporting. The AI is a real and heavily marketed layer on a product that stands without it. Checked 7 September 2026.
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 here rather than asserted, and it is documented, but nothing measures it. Every fact in the chronology is linked to the source it came from, evidence citations are included on every plan including the non-AI tier, the assistant is described as returning answers with clickable citations linking directly to the user's source documents, and Proposed Facts arrive with a citation back to the source for each drafted fact.
That is linked source material the reader can open and check, which is what the middle band asks for, and the retrieval target is the case record rather than published legal authority. The agreement is unusually candid about the limits: the AI Features section states that the features may not be 100 per cent accurate and may contain errors, omissions or misinterpretations, that Casefleet makes no representations or warranties regarding the accuracy, completeness or reliability of any AI-generated output, and that beta or preview features should be used with heightened caution regarding accuracy and reliability.
What is absent is measurement of any kind: no accuracy figure, no test set, no evaluation, no error taxonomy and no described retrieval method beyond the citation link itself. Grounding to primary legal authority and citation status checking do not bite on a fact management product and are counted neither way. Checked 7 September 2026.
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.
A written commitment to supervision plus a real review surface, short of the full control structure. The review surface is a hard gate rather than a policy statement: Proposed Facts are drafted by Casey and staged for the user, who reviews, edits and approves every one before it joins the chronology, and the product page for the assistant describes Casey setting up cases, proposing facts and organising files while the user approves every step.
The agreement reinforces it contractually, stating that AI-generated content and analyses are tools to supplement, not replace, professional legal work and human oversight, and that the user is solely responsible for reviewing, validating and verifying any output before relying on it. An activity feed records every change across a case and the platform applies audit logging. What holds this below the top band is the boundary around the autonomous mode.
Agentic Workflows are marketed as handing Casey multi-step case tasks, and nothing published states what an agent may do unattended, at what point it stops, what a user sees while it runs, or what happens after it acts wrongly. No confidence signal is described as surfaced to the reviewer. The agentic workflows feature page was not opened in this pass and is named as the artifact that could move this row; it is amendable on that evidence. Checked 7 September 2026.
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.
Testimonials stand in for deployment evidence and no figures accompany them. Six customer quotations are published. Three carry a person's name and role but no organisation: Jason Shinn and Paul Fleming, both described only as Attorney, and Risa Beck, Paralegal Manager. The other three are attributed by state and role alone, being a South Carolina consumer protection lawyer, an Arkansas investigative paralegal and an Illinois criminal law attorney.
No firm or legal department is named anywhere on the surfaces read, no customer logos appear, no case study is published, and no operational figure of any kind is offered: no customer count, no adoption number, no time saved, no outcome measured. The one durable claim is longevity, that the company has been helping to win cases since 2016, which is a statement about the vendor rather than about a deployment. The quotations themselves are substantive and read as genuine practitioner comment, including one describing the chat feature as becoming a go-to legal AI, but a reader cannot assess any of it because no basis is stated. This grade records the shape of the evidence rather than doubt about the customers. Checked 7 September 2026.
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.
Substantive published commitments, several of them AI-specific, short of vendor-side privilege treatment. The agreement grants Casefleet rights over customer data only for the purpose of providing the service and states expressly that it is for no other purpose. A confidentiality section binds both parties, and it carries an AI-specific clause: Casefleet states it implements technical and organisational measures to protect confidentiality when data is processed through AI Features, that its AI systems are designed to respect the confidential nature of legal work, and that it implements safeguards to prevent commingling of confidential data between different users and organisations.
Segregation is supported in the product by granular administrator control over which data is visible to which users, described on the security page as ensuring privileged and confidential information stays that way. Data is anonymised by UUID between infrastructure components, files are virus scanned, and on account deletion Casefleet undertakes to delete or destroy all copies of customer data in its possession unless legally prohibited.
Privilege and work product are named in the agreement, but as the user's compliance obligation rather than as a statement of how Casefleet handles privileged material, so the top limb is not met. One provision cuts the other way and belongs on the record: the warranty section states that the user bears sole responsibility for adequate security, protection and backup of their data, that Casefleet does not guarantee transmissions to third-party vendors will be encrypted, and that Casefleet has no liability for unauthorised access to or loss of customer data. Checked 7 September 2026.
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.
A real published position on advice versus tooling, developed further than most records carry, held below the top band on two limbs. The agreement devotes a dedicated section to the question, headed No Legal Advice and No Attorney-Client Relationship, stating that Casefleet is not a law firm, that no attorney-client relationship arises from use of the platform including its AI Features, and that the platform and its AI are software tools rather than a substitute for the advice of an attorney.
Five numbered acknowledgements follow, and two of them address non-lawyer users directly: a user who is not a licensed attorney should consult one before making decisions based on information obtained through the platform, and a user operating in a capacity other than as a licensed attorney bears sole responsibility for obtaining appropriate legal counsel. The competence and supervision dimension is present too, with the AI Features section requiring the user to ensure their use complies with professional ethical obligations, naming attorney-client privilege, the work product doctrine and confidentiality requirements.
That treatment is deliberate and it matches a real exposure, because the vendor publishes use-case pages aimed at people handling their own matters, including VA disability claims, disability appeals, small claims preparation, custody documentation, insurance claim disputes and estate settlement. Two limbs are not established. Jurisdiction limits are addressed only obliquely, through an acknowledgement that AI output may be inapplicable to the user's jurisdiction, rather than by naming where the product is intended to be used.
And whether those consumer-facing pages themselves carry a clear disclosure could not be established: none was opened in this pass, and they are named as the artifact that would settle the top band. Checked 7 September 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.
No governance disclosure was located, and the grade should be read against what the vendor does publish so the absence is not mistaken for general silence. The agreement does substantial AI work: it defines AI Features as a contractual term, devotes a section to their limitations, allocates ownership of AI output, and commits on training. All of that is disclosure about the product and about data handling, and it is credited on the axes where it belongs.
None of it is governance. Nothing published describes an AI policy, a responsible AI or ethics position, a bias or fairness assessment, model evaluation or testing of any kind, a named internal owner for AI decisions, or alignment to any governance framework such as ISO 42001 or the NIST AI Risk Management Framework. Nothing addresses how models are selected, assessed before release, or monitored in use, on a product whose agentic features act on litigation case files.
Surfaces read on 7 September 2026: the homepage and full features and use-case navigation, the Terms of Service in full, the pricing page and its FAQ, and the security page.
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.
Stewardship is contractual and AI-specific, which is rare, and two gaps hold it below the top band. The agreement commits that Casefleet will not use customer data to train its AI systems without the customer's explicit consent, and separately states that Casefleet maintains a closed-loop system in which customer data is not used to train AI models. The data licence is purpose-limited in express terms, granted only to provide the service and for no other purpose.
Ownership of AI output is allocated to the customer to the extent it derives from their data, with Casefleet making no claim to it while retaining rights in the underlying technology. The confidentiality section adds AI-specific technical and organisational measures and a stated safeguard against commingling data between users and organisations. Security controls are described in unusual operational detail, covering encryption in transit and at rest, a virtual private cloud, virus and malware scanning of all uploads, daily backups, least-privilege employee access with revocation on termination, code review before release, and a published vulnerability disclosure route.
Two things hold it at B. The agreement discloses that AI Features process data through third-party AI services without naming any of them, so the buyer cannot see who is in the loop. And the warranty section undercuts part of the security position by stating that Casefleet does not guarantee that transmissions to its third-party vendors and hosting partners will be encrypted, and disclaims liability for unauthorised access to customer data. Checked 7 September 2026.
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.
A published agreement with a stated cap and a real vendor-side indemnity, and no recourse at all for AI output. On the credit side, the liability cap is quantified rather than formulaic, limited to fees actually paid in the six months preceding the claim or five hundred dollars where no fees apply, with indirect and consequential damages excluded. Casefleet gives an intellectual property indemnity running in the customer's favour, undertaking at its own expense to defend infringement actions arising from use of the software within scope, with a right to procure, replace or modify, and a pro-rata refund if it cannot.
Price changes carry 30 days' notice, charges may be disputed within 60 days, and data is available for download for 30 days after termination. Against that, the AI-specific position is a complete disclaimer: the agreement states that Casefleet makes no representations or warranties regarding the accuracy, completeness or reliability of any AI-generated output, that the user is solely responsible for reviewing and verifying it, and that no output constitutes legal advice.
The service is provided as is and as available with all warranties disclaimed, there is no uptime commitment, no service credit and no cure period, and Casefleet disclaims liability for corruption, deletion or loss of customer data. A firm whose chronology is built on a wrongly extracted fact has a contractual remedy of nothing, capped at six months of fees in any event. Checked 7 September 2026.
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.
No integration with any legal practice system was located on any surface read. The site has no integrations page, no partner listing and no developer or API documentation anywhere in its navigation, which covers features, use cases, pricing, resources and support. No document management system, practice management system, case management system, billing system, e-discovery platform or court filing system is named anywhere.
The agreement acknowledges the category in the abstract, defining Non-Casefleet Products as third-party applications a customer may use in connection with the platform whether or not Casefleet designates them official integrations, which establishes that the concept exists while naming nothing that occupies it. The agreement runs the other way on programmatic access, prohibiting users from using any non-public Casefleet APIs.
Single sign-on via SAML2 is offered on Enterprise plans, but that is authentication rather than practice-system integration and it is credited on security. The product is designed as a self-contained workspace, which is a coherent choice for a chronology tool, and this grade records the disclosure position rather than a defect: a buyer evaluating how Casefleet would sit alongside their existing stack has nothing published to work from. Checked 7 September 2026.
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.
Neither limb is stated, which is what the band requires. The hosting provider is named plainly and repeatedly: all Casefleet infrastructure is hosted on Amazon Web Services and all customer data is stored there, with the security page linking to AWS's own security and data centre control documentation. Architecture is described at a level most vendors do not attempt, with infrastructure handling end-user data living entirely within a virtual private network engineered to expose only required components to the public internet, all other components reachable only over the private network, encryption required for internal as well as external communication, and daily backups retained for an extended period.
None of that answers either question this axis asks. No region or data centre location is stated anywhere, no residency option or commitment is offered, and nothing addresses where data sits for a customer outside the United States. Tenancy is equally absent: no statement describes whether the service is single-tenant, pooled or configurable, and the anonymisation of data by UUID between components is an internal design detail rather than a tenancy model.
No self-hosted or private deployment option was located. Naming the hosting provider identifies who runs the infrastructure, not where the data rests. Checked 7 September 2026.
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.
Controls are described in detail and no certification of any kind exists. The security page is substantive and specific across five headings, covering application security with XSS and CSRF protection, signed cookie session management, rate limiting, audit logging, password hashing and malware scanning; network and storage security with TLS required internally and externally and encryption at rest; operations with automated security testing, mandatory peer code review, staging before release, daily backups and disk-encrypted company laptops with firewalls and network monitoring for engineers holding production access; account security with two-factor authentication, SAML2 single sign-on on Enterprise, granular permission levels and session tracking that alerts on concurrent logins; and physical security through AWS and key card office access.
A vulnerability disclosure route is published with a stated address, a request for reproduction detail and an ask for reasonable time before public disclosure, which is a responsible disclosure posture rather than a certification. What does not exist anywhere is third-party attestation. No SOC 2, no ISO 27001, no certification of any kind, no auditor or certification body named, no penetration test disclosed, and no trust centre or document portal.
The claim that every piece of code, infrastructure and operations is continually audited for security impact is an internal practice statement, not an external audit. HIPAA is addressed commercially rather than by certification, with the vendor stating it is HIPAA compatible and will sign a business associate agreement on annual accounts for an additional fee. Nothing unsupported is claimed, which is why this does not sit lower. Checked 7 September 2026.
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 shape of the supply chain is disclosed and nothing in it is identified, which is an unusual combination worth stating precisely. The agreement tells a buyer more than silence would: it states that AI Features utilise multiple AI models and technologies from various providers to deliver functionality and that specific features may be powered by different underlying systems, and separately that when AI Features are used, data may be processed through various artificial intelligence systems including third-party AI services.
A reader therefore learns that this is a multi-provider architecture, that providers are external, and that the model behind one feature may differ from the model behind another. That is an honest description of the structure. What follows none of it is a name. No model, no version, no family, no provider and no infrastructure route for inference is identified on any surface read, and there is no subprocessor list, DPA or trust centre where such a list might otherwise sit.
Amazon Web Services is named as the hosting provider for all Casefleet infrastructure, which says where the application runs rather than whose model reads a client's case file, and it has been credited on deployment rather than counted a second time here. The result is that a firm can tell its client that several unnamed third parties process case material through AI, which is a disclosure that raises the question it does not answer. Checked 7 September 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.
This is the most granular pricing disclosure located in the corpus, and it lands in the middle band because the top of the range is withheld. What is published without any sales contact: Starter at 30 dollars per user per month, Advanced AI at 140 dollars per user per month, a 22 per cent discount for annual payment, and a full plan comparison table setting out what each tier includes down to document limits of 20 per case on Starter against no limit above it.
The unit of charge is stated for every metered component, and so are the overage rates: optical character recognition at one dollar per hundred pages after 3,000 free pages per billed user per month, media transcription at six cents per minute after three free hours, document intelligence at one cent per page after 12,000 free pages, storage at ten dollars per gigabyte per month after five gigabytes included, and AI credits at one tenth of a cent per credit after 100,000 free credits per billed user per month.
Publishing an AI metering rate to that precision is rare enough to note. The agreement adds the billing mechanics: credits are allocated monthly, do not roll over, vary in consumption by feature, are non-refundable, and can be capped by the customer in settings, and AI Features are available only on usage-based billing. The trial is 14 days with no credit card and a stated 10-document-per-case limit. Two things sit against the top band.
Enterprise, for twenty or more users, is a custom quote. And the additions a buyer would meet on implementation are named as categories without figures: the HIPAA business associate agreement carries an unstated additional fee, bulk uploads above 1,000 files a day are quoted upfront, and Enterprise data migration is unpriced. Checked 7 September 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.
Practice coverage is genuinely broad and set out page by page rather than asserted. Dedicated practice-area pages cover employment law, criminal law, personal injury, workers' compensation, family law, complex litigation and medical chronologies. Workflow pages cover case management, litigation management, investigations, cross-examination preparation, summary judgment and early case assessment, which is coverage by litigation stage rather than by subject matter alone and is the more useful cut for this product.
The vendor states it serves both plaintiff and defence firms, and two free tools support practice work directly: a litigation deadline calculator and a statute of limitations lookup organised by state and claim type. A further set of pages addresses individuals handling their own matters, covering VA disability claims, disability appeals, small claims, custody documentation, insurance disputes and estate settlement, which widens the audience beyond law firms.
What holds this at B is that the jurisdiction is single and the firm dimension is thin. Everything is United States practice, with the agreement governed by Georgia law and Fulton County courts holding exclusive jurisdiction, and no international coverage is claimed. On firm size, the only published signal is that Enterprise begins at twenty users; no segment statement, firm-size breakdown or named practice is published. Checked 7 September 2026.
From $30 per user per month
- Casefleet publishes its full price list. Starter is $30 per user per month and carries the case chronology, document reviewer and collaboration tools with a 20-document limit per case and no AI. Advanced AI is $140 per user per month and adds the Casey assistant, document intelligence and unlimited documents. Enterprise, for twenty or more users, is a custom quote. Paying annually saves 22 per cent. Metered extras are priced per unit, including AI credits at $0.001 each after 100,000 free credits per user per month, and a 14-day trial needs no credit card.
Published rates with the unit of charge stated for every metered component, and only the enterprise tier withheld. Plan rates: Starter $30 per user per month; Advanced AI $140 per user per month; Enterprise a custom quote for twenty or more users. Annual payment saves 22 per cent. A full comparison table sets out what separates the tiers, including a 20-document-per-case limit on Starter against no limit above it, full-text search excluded from Starter, and the entire document intelligence and Casey assistant feature sets available only from Advanced AI upward.
Metered usage is priced to the unit and published on the same page: optical character recognition $1 per 100 pages after 3,000 free pages per billed user per month; media file transcription $0.06 per minute after 3 free hours per billed user per month; document intelligence $0.01 per page after 12,000 free pages per billed user per month; AI credits $0.001 per credit after 100,000 free credits per billed user per month; storage $10 per GB per month with 5 GB included on Advanced AI and Enterprise.
The agreement adds the billing mechanics: AI credits are allocated monthly on billed users, do not roll over, are consumed at different rates by feature, are non-refundable including overages, and can be capped by the customer in account settings, with AI Features available only to accounts on usage-based billing rather than the legacy plans that predate 1 December 2024. Trial terms are stated: 14 days, no credit card, limited to 10 documents per case.
Charges are per user, disputes must be raised within 60 days, price changes carry 30 days' notice, and past due amounts attract 1.5 per cent monthly. What is not priced: the Enterprise tier, the additional fee for a HIPAA business associate agreement, bulk uploads above 1,000 files per day which are quoted upfront, and Enterprise data migration. Surfaces read on 7 September 2026: the pricing page including the comparison table and FAQ, and the Terms of Service in full.
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?
Training occurs only where the customer has affirmatively enabled it.
The agreement addresses training expressly and structures it as consent-gated, which is why this takes the opt-in value rather than a never value. The operative words are that Casefleet will not use customer data to train its AI systems without the customer's explicit consent. A consent route therefore exists, and a value asserting the vendor never trains would misstate the agreement. Alongside it, in the AI Processing paragraph, the agreement states that Casefleet maintains a closed-loop system in which customer data is not used to train AI models, which describes the current architecture in unqualified terms.
Both statements sit in the same instrument and the honest reading is that the default is no training, the architecture is described as closed-loop, and a consented departure is contemplated. Nothing published describes how consent would be sought, at what granularity, or whether it can be withdrawn. Two qualifiers belong on the record. The agreement separately reserves the right to use anonymised and aggregated data derived from user interactions with AI Features to improve the platform, and elsewhere to generate and publish aggregate anonymised reports on system usage and content trends.
Those operate on de-identified interaction and usage data rather than on case content, and neither is read as a training right, but a buyer weighing the closed-loop statement should see them. The data licence itself is purpose-limited, granted only to provide the service and, in the agreement's own words, for no other purpose.
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.
Nothing addresses how long prompts to the assistant or the outputs it returns are retained. The agreement sets an account-level rule on exit: after cancellation or termination, data is available for download for 30 days and is then deleted, and if a team account is deleted Casefleet undertakes to delete or destroy all copies in its possession in a reasonably expedient way unless legally prohibited. That governs stored case material, which would include facts once approved into a chronology, and it says nothing about the AI interaction layer: no period covers a query put to Casey, a draft fact that was rejected rather than approved, a summary generated and discarded, or the intermediate output of an agentic workflow.
Nothing states whether prompts persist at all, whether they are retained by the unnamed third-party AI services the agreement discloses, or whether a customer can purge them. Usage is metered at the account level for billing, which establishes that interaction records exist without stating what is kept or for how long. Surfaces read on 7 September 2026: the Terms of Service in full, the pricing page and FAQ, the security page, and the AI feature descriptions in the site navigation.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
The product maintains its own permission model, documented, requiring the firm to keep it aligned.
A separation model is documented at two levels and it reaches the mechanism a firm would actually use. Between organisations, the confidentiality section commits Casefleet to implement safeguards preventing the commingling of confidential data between different users and organisations, and states this specifically in relation to processing through AI Features, which is the point at which commingling risk is least visible to a customer.
Within a firm, the security page describes granular administrator control over which data is visible to which users, framed expressly as ensuring privileged and confidential information stays that way while still allowing collaborators to be invited. That is the operative capability behind an ethical wall: an administrator can exclude a conflicted user from a case. Supporting controls are described, including anonymisation of data by UUID so each infrastructure component knows only the minimum required, session tracking with alerts on concurrent logins, and two-factor authentication.
What is not published is the conflicts dimension proper. Nothing describes a conflicts check, a screening workflow, a record that a wall was erected and when, or whether the AI features respect user-level permissions when the assistant searches across a case. The mechanism exists and is documented; its use as a compliance artifact is not.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
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.
Compelled disclosure is addressed squarely in two separate clauses and neither carries any notice commitment. The confidentiality section provides that both parties reserve the right to provide confidential information to third parties as required and permitted by law, giving a subpoena or court order as the example, and to cooperate with law enforcement authorities in the investigation of any criminal or civil matter.
Separately, the representations section has the user acknowledge, consent and agree that Casefleet may access, preserve and disclose account information and customer data where required by law or in a good faith belief that doing so is reasonably necessary to comply with legal process, enforce the terms, respond to support requests, or protect the rights, property or personal safety of Casefleet, its users and the public.
The second is broader than compelled disclosure, since a good faith belief that disclosure is reasonably necessary is a lower threshold than a legal requirement. Across both, nothing commits Casefleet to notify the customer, to resist or narrow a demand, to disclose only the minimum required, or to keep a record of what was disclosed, and no transparency report is published. The value records that the vendor engages with the question and stops short of the commitment.
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.
Two questions sit here and only one applies to this product. Casey operates on the customer's own case file, being the documents, transcripts and media a firm uploads to a matter, so there is no external legal corpus, case law library or third-party content set whose sourcing and licensing a buyer would need to check. That limb does not apply and is not counted against the vendor. The limb that does apply is the provenance of the models themselves, and nothing addresses it.
The agreement discloses that AI Features utilise multiple models and technologies from various providers and that data may be processed through third-party AI services, without naming any of them, so a buyer cannot establish what any underlying model was trained on, on what basis, or under what licence Casefleet uses it. The one adjacent published artifact is the statute of limitations lookup, a free tool presenting limitation periods by state and claim type, whose own sourcing is not stated; it is a marketing tool rather than part of the graded product and is named here only so a reader knows it was seen.
Surfaces read on 7 September 2026: the Terms of Service in full, the security page, the pricing page and the site navigation.
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.
The product ships no citator and makes no good-law claim, which is the expected position for case chronology and fact management software. Casey's citations point to the customer's own source documents rather than to legal authority, so there is no published case to validate, no treatment signal and no subsequent history to check. Two free tools on the site touch legal currency without being citators: a litigation deadline calculator and a statute of limitations lookup by state and claim type, both of which present rules rather than verify authority, and neither is part of the platform a customer buys.
Recorded at the floor because the value set requires a value, with the reason stated here so a reader does not take it as a finding against the vendor. Nothing else in this record depends on it.
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.
Nothing describes how the system behaves when it is unsure. The agreement is candid that uncertainty exists, stating that AI Features may not be 100 per cent accurate and may contain errors, omissions or misinterpretations, and that beta or preview features should be used with heightened caution regarding accuracy and reliability. That is an honest admission of the condition rather than an account of the behaviour: it tells a user to expect error without telling them what the system does when it encounters one.
No confidence score, likelihood indicator or reliability signal is described as surfaced to the reviewer, nothing states when Casey declines to answer or flags a low-confidence result, and nothing addresses what happens with a document it cannot parse or a question the case file does not support. The Proposed Facts approval gate is the nearest surface and it is named rather than credited: staging every drafted fact for human approval is an oversight control that applies uniformly, not a behaviour that varies with the model's own uncertainty, and it is graded on autonomy where it belongs.
Surfaces read on 7 September 2026: the Terms of Service in full, the homepage AI descriptions, the pricing page and the security page.
Fabricated Citation Record
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.
No matter naming Casefleet or LAWPRCT, Inc. was located in the hallucination case tracking maintained by Damien Charlotin or in the sanctions reporting drawn from it, searched on 7 September 2026 on both the product name and the registered corporate name. The tracked corpus is now substantial and reporting on it names the tools involved where they are known, with the database logging well over a thousand matters worldwide and United States sanctions ranging from four-figure fines to licence suspension.
Casefleet appears in none of it. The product class is relevant to why: Casey cites to the customer's own uploaded documents rather than generating citations to legal authority, and every drafted fact is staged for human approval before entering a chronology, so the specific failure mode this signal tracks is structurally unlikely rather than merely unrealised. Recorded as none located rather than as a positive finding about the vendor.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Public materials refer to professional responsibility in general terms without naming guidance.
Professional obligations are referenced in general terms without any authority being named. The agreement requires the user to ensure that their use of AI Features complies with all applicable laws, regulations and professional ethical obligations, and it names the categories those obligations fall into, being attorney-client privilege, the work product doctrine, confidentiality requirements and data protection laws.
A separate section states that Casefleet is not a law firm, that no attorney-client relationship arises, and that a user who is not a licensed attorney should consult one. That is a real engagement with the ethical frame and it is why this sits above the floor. What is absent is any named source of guidance. No bar association, state bar opinion, ABA model rule, court standing order or regulator publication on the use of artificial intelligence is cited, referenced or mapped anywhere, and nothing is broken down by jurisdiction despite the product being sold across United States practice with a state-by-state limitation tool on the same site.
A firm looking for help aligning its own AI use to its regulator's expectations finds an instruction to comply rather than a route to the guidance.
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, and the product sits inside a fee relationship between a lawyer and a client where those savings would change the bill.
Savings are claimed and the fee consequence is nowhere addressed, with an unusual twist worth recording. The savings claims are the vendor's own rather than a customer's, framed as cutting hours of manual reading, reducing repetitive manual tasks and reviewing faster, with no figure attached to any of them. Nothing addresses what happens to a client bill when document review that took a day takes an hour, no guidance on fee or disclosure treatment is offered, and no per matter record of AI-assisted work is described.
The twist is that Casefleet meters AI work more precisely than almost any vendor in this index, allocating credits monthly, consuming them at different rates by feature, publishing an overage rate per credit, and making usage reports and billing detail available to account administrators. The metering exists for the vendor's own billing, and it is pitched at the account, not the matter: nothing ties credit consumption to a case, so a firm holding a precise record of how much AI it bought cannot derive from it how much AI went into any particular client's file.
The raw material for a per matter record is closer to hand here than in most records, and it has not been assembled into one.
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.
A firm cannot answer a client's AI clause from anything Casefleet publishes. No subprocessor list exists on any surface read, and there is no data processing agreement, trust centre or document portal where one might be requested. No model provider is named. The agreement discloses the opposite of what a client wants to hear, stating that AI Features utilise multiple models and technologies from various providers, that specific features may be powered by different underlying systems, and that data may be processed through third-party AI services, all without naming a single one.
A firm asked whose model reads its client's medical records or deposition transcripts could report only that several unnamed third parties are involved. Amazon Web Services is named as the host of all infrastructure and data, but infrastructure identifies where software runs rather than whose model sees client content and does not satisfy this signal. No forwardable client-facing material was located: no consent or notification pack, no client disclosure template and no published processing terms.
Two adjacent artifacts are named and set aside because they answer different questions: a business associate agreement is offered for HIPAA on annual accounts for an additional fee, and single sign-on is available on Enterprise. Surfaces read on 7 September 2026: the Terms of Service in full, the security page, the pricing page and FAQ, and the complete site navigation.
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.
The record this product creates is exportable and litigation-facing by design, and it is not framed as a record of AI involvement. What exists is substantial: a chronology in which every fact carries a link to its source, evidence citations on every plan, litigation-ready reports generated in a click, case outlines tying issues and arguments to evidence, and an activity feed described as showing every change across a case, supported by application-level audit logging.
A firm asked to show how a fact was established has the source link to produce, which is the underlying discipline this signal cares about. What is not established is whether that record identifies the AI's part. The Proposed Facts workflow means every AI-drafted fact passes through a human approval step before entering the chronology, so a distinction between proposed and approved plainly exists inside the product, but nothing published states whether the activity feed or any export preserves it, whether a report can show which facts originated with Casey, or whether an agentic workflow's actions are logged in a form a court could read.
Nothing anywhere addresses disclosing AI use to a court or tribunal, and no certification, template or guidance is offered. The record is real and usable and partial for this purpose.