CARET Legal

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 models are the engine of a core capability layered on a product that would still function without them. CARET AI finds unbilled time in work already logged, flags what needs attention and calculates deadlines from matter data, and Quick Summary condenses documents; remove them and case management, native email, workflows, document management, billing, trust accounting and payments remain, which is the product the company has sold for forty years. The terms treat AI Services as a defined subset of the subscription, carve them out of the service-level commitment and warranty, and price AI summaries on the second tier upward. Home page, features page, pricing page and terms read 6 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.

Accuracy is disclaimed without measurement and the primary-authority limbs do not apply. Terms section 20.3 states that AI outputs may be inaccurate, that CARET makes no representation as to their accuracy or reliability, and that every output must be independently reviewed before use; the 2023 launch release claims Quick Summary accelerates decisions without sacrificing accuracy, with no test set, figure or evaluation published. The features summarise documents and surface matter data rather than citing law, so the citation limbs are not held against the product; what the band asks of a summariser, a measure of how often summaries are wrong, is not published. Terms, launch release and blog posts read 6 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 that the models work alongside a reviewing person, short of the full control structure. Terms section 20.3 provides that outputs are for informational and workflow-assistance purposes, must be subject to independent review and validation before use, and must not be relied on without appropriate human review by qualified personnel, and requires the client to report inaccurate or harmful results to CARET; CARET AI presents unbilled time, attention items and deadlines for a person to act on. Nothing states a threshold at which the system acts alone, what executes without approval, or a route back after a wrong summary or deadline beyond the client's own correction, and section 23 places responsibility for missed deadlines and sanctions on the client. Terms and home page read 6 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.

Scale claims and an illustrative estimate stand in for named deployments on the surfaces read. The home page states more than 6,500 firms and forty years in the market, and the pricing page claims small and mid-sized firms can capture an extra twenty thousand dollars of billable work per timekeeper, footnoted as an informational estimate that may vary significantly. No named customer with a figure was located; the customer stories page exists in the navigation and was not opened, and is the rebuttal route. Home page and pricing page read 6 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.

Confidentiality is addressed in the agreement and the training position is adverse, which is the finding. Terms section 9 treats client data as confidential with a protective-order notice clause on legal process, section 7 deletes client data thirty days after the term, and the 2023 and 2025 blog posts state that Quick Summary runs in a closed environment so client data is not exposed outside the firm's instance. Against that, section 20.4 licenses the client's AI inputs and outputs to CARET to train, tune and evaluate models and features, with learnings retained beyond the term as anonymised aggregated data, and section 21.5 permits the same for beta data; the agreement governs over the marketing. No treatment of privilege or work product and no description of matter-level segregation was located. Terms and blog posts read 6 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 position on advice versus tooling, framed for a professional user, short of a competence dimension and with a jurisdiction statement that shifts the burden. Terms section 20.3 states that AI outputs do not constitute legal, accounting, tax or professional advice, require human review by qualified personnel, and that the client is responsible for legally compliant AI disclosures to its own clients; section 23 states that CARET does not warrant compliance with court rules or deadlines in any jurisdiction and that the client must determine whether its use complies with the rules of every jurisdiction it is subject to, and disclaims liability for missed deadlines, sanctions and malpractice claims. The buyer is a law firm. No bar guidance is named. Terms read 6 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.

Responsible-use statements without a mechanism, testing regime or accountable owner. The 2025 blog post states that CARET regularly reviews how it uses or plans to use AI and the tools that support it, and the terms bar using AI services to generate unlawful, harmful or discriminatory content; no governance framework, ISO 42001 or equivalent, pre-release testing description or statement about uneven output is published on the surfaces read. Blog posts, terms and security navigation read 6 September 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.

Some of the ground is covered in the agreement and the rest sits on surfaces not read. Retention and deletion: terms section 7 permanently deletes client data from all CARET and affiliate storage thirty days after the term ends. Sub-processors: section 37 permits subcontracting without consent and section 34 incorporates a data processing addendum by reference, which was not opened; Azure OpenAI is named as the AI provider on the blog. Access control and incident practice: the data security page exists in the navigation and was not opened. Section 27 permits disclosure or transfer of client data in a merger or sale. Terms and blog posts read 6 September 2026; the security page and DPA are the rebuttal route.

AI Liability and Recourse

What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.

What the vendor stands behind is published and specific, including that on wrong output it stands behind nothing, which a buyer can read before signing. Terms section 25 gives a CARET defence and indemnity for third-party claims that the subscription service infringes intellectual property, with exclusions and the procure, modify or terminate remedies; section 24 caps CARET's aggregate liability at subscription fees paid for the service in the six months preceding the claim and excludes indirect and consequential loss; section 23 warrants only the service-level commitment, excluding AI services, with a service credit as sole remedy, disclaims all warranties on AI outputs and states that CARET is not liable for missed deadlines, late filings, sanctions, adverse rulings or malpractice claims arising from reliance on the service; section 30 bars actions after two years. The client indemnifies CARET for AI inputs. No insurance is stated. Terms of service of 18 August 2026 read in full 6 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.

Integrations are referred to without documentation an implementer could use on the surfaces read. The terms contemplate APIs and third-party applications subject to rate limits and a per-seat rule, disclaim function when combined with third-party software including Microsoft products, and the platform includes native email and calendaring; an integrations page exists in the navigation and was not opened, and no integration is described with what moves and in which direction. Terms and navigation read 6 September 2026; the integrations page is the rebuttal route.

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.

Cloud delivery is stated and neither tenancy nor region is addressed with substance. The terms define a SaaS-based subscription service, and the blog posts state that Quick Summary runs in a closed environment so that client data is not exposed outside a firm's instance of CARET Legal, which implies per-firm separation without saying whether infrastructure is shared or where it sits. No region, hosting provider or residency option is stated on the surfaces read; the data security page was not opened and is the rebuttal route. Terms and blog posts read 6 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.

No certification with scope was located on the surfaces read. The terms carry no security warranty beyond confidentiality obligations, a status page exists, and a data security page sits in the navigation and was not opened; no SOC 2, ISO 27001, auditor or report route appears in the material read. This grade is rebuttable on the security page. Terms and navigation read 6 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 supply chain is partly disclosed. The 2023 launch release and the 2025 blog post state that Quick Summary is powered by Microsoft's Azure OpenAI Service in a closed environment, which names the provider and the hosting arrangement; the terms acknowledge third-party and publicly available models and state that CARET's providers retain rights in the underlying models. What is not published is the model itself, the providers behind the 2026 CARET AI features, where inference runs, or a change-notification commitment. Launch release, blog post and terms read 6 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.

A figure, a unit and a term are published. The pricing page lists Enterprise at 79, Enterprise Plus at 99 and Enterprise Insights at 119 dollars per user per month billed annually, plus one-time implementation fees, with the features of each tier itemised and AI-generated document summaries starting on Enterprise Plus; premium support is an add-on with a 25-user minimum. The terms add that AI services may carry additional fees and token or credit limits set in the order, that fees may rise annually or on thirty days' notice, and that termination before term end incurs the remaining balance. Pricing page and terms read 6 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.

Segment and practice coverage are described with substance and some limits are stated. The vendor publishes solution pages for twelve practice areas including litigation, personal injury, estate planning, intellectual property, criminal, immigration, family, real estate, employment, insurance defence and workers' compensation, role pages for administrators, partners, bookkeepers and IT, and firm-size pages for one to thirty and thirty-plus employees; the terms state the service is US-governed and subject to export control, and AI features are limited to the second tier and above. No practice area is named as unsupported. Navigation, pricing page and terms read 6 September 2026.