C

CARET Legal

The AI drives a core capability on a practice management product that ran without it for decades. CARET AI finds unbilled time in work already logged, flags matters that need attention and calculates deadlines from matter data, and Quick Summary condenses documents; the company introduced more AI in its 2026 Summer Launch. Remove those features and case management, native email and calendaring, automated workflows, document management, reporting, billing, trust accounting and payments remain, which is the product CARET and its predecessor AbacusLaw have sold to firms for some forty years.

The Terms of Service treat AI Services as a defined subset of the subscription: they are excluded from the 99.9 percent availability commitment and from the warranty, may carry separate fees and token or credit limits set in the order, and document summaries start on the second pricing tier. In that framing the vendor states plainly that the product stands without the models.

Accuracy is disclaimed in the contract, asserted in marketing and never measured. CARET's AI does not cite law; it summarizes documents, surfaces unbilled time and calculates deadlines. The question is how often a summary misstates a document, a time entry is wrong or a computed deadline is off, and the last matters most because a missed deadline is a malpractice claim. Section 20.3 of the Terms says AI outputs may be inaccurate, that CARET makes no representation about their accuracy or reliability, and that every output must be independently reviewed before use.

The 2023 launch announcement claimed Quick Summary speeds decisions without sacrificing accuracy, with no test, figure or evaluation behind it. For deadlines, CARET also integrates LawToolBox, which calculates dates from the rules of civil procedure and tracks rule changes; nothing says whether CARET AI's own deadline calculations use it. Section 23 disclaims liability for missed deadlines and late filings.

A written commitment that people review model output, without the full control structure. Section 20.3 of the Terms of Service says AI outputs are for information and workflow assistance, must be independently reviewed and validated before use, and must not be relied on without appropriate review by qualified personnel. It asks the firm to report inaccurate or harmful results to CARET, a feedback route. In practice CARET AI presents unbilled time, attention items and calculated deadlines for a person to accept and act on rather than posting them itself, and Quick Summary produces a summary for a user to read.

Nothing says where automation stops: whether a calculated deadline goes onto the calendar without confirmation, whether unbilled time suggestions can post to an invoice on their own, or what happens after a wrong output is reported beyond the firm's own correction. Section 23 places responsibility for missed deadlines and sanctions on the firm.

Named firms report specific results, though none measures the AI features. The customer stories page profiles named firms and people. At Tectonic, LLP, a San Francisco IP boutique, attorney Hannah Stitt says custom intake forms and a task template cut the time from lead to signed retainer to 2.5 hours. At LaMagna McKenna, partner Michael La Magna estimates 15 to 20 hours a month saved after a merger. Other stories cover Diller & Rice, LLC, a three office Ohio firm; Pregenzer Baysinger Wideman & Sale, PC, whose firm administrator describes leaving a legacy system with frequent downtime; Doering, Grisell & Cunningham, P.A. in Kansas, which moved from CARET's on premises AbacusLaw to the cloud; and Parks Zeigler in Virginia Beach.

CARET's home page says more than 6,500 firms use it, and the pricing page's estimate of an extra $20,000 of billable work per timekeeper is footnoted as an informational estimate. The figures are the customers' own, are not dated, and none is attributed to CARET AI.

Confidentiality is covered in the agreement, and the agreement also licenses the firm's AI inputs and outputs for model training. Section 20.4 of the Terms of Service gives CARET a license during the term to use AI inputs and outputs to provide and improve the AI Services, to train, tune and evaluate models and features, and for internal analytics. Anything kept beyond the term is held as anonymized aggregated data, and section 21.5 does the same for beta data.

That contradicts CARET's 2023 and 2025 blog posts, which say Quick Summary runs in a closed environment on Azure OpenAI so client data is not exposed outside the firm's instance; where the two conflict, the agreement governs. Section 9 treats client data as confidential and requires notice so a firm can seek a protective order before compelled disclosure, and section 7 deletes client data from all CARET storage 30 days after the term ends. Attorney client privilege, work product and separation between matters are not addressed.

A real position on advice versus tooling is framed for a professional user, with jurisdictional compliance placed on the firm. Section 20.3 of the Terms of Service says AI outputs are not legal, accounting, tax or professional advice and need review by qualified personnel, and that the firm is responsible for any legally required disclosure of AI use to its own clients. Section 23 says CARET does not warrant compliance with court rules or deadlines in any jurisdiction and that the firm must decide whether its use complies with the rules of each jurisdiction it is subject to.

CARET is not liable for missed deadlines, sanctions, adverse rulings or malpractice claims arising from reliance on the service. The buyer is always a law firm, the AI features serve the firm's own staff rather than the public, and the Terms bar using the AI for unlawful or discriminatory content. No bar opinion or regulator guidance on lawyers' use of AI is cited, and nothing links AI generated time entries to a lawyer's duty to bill accurately.

There are statements on responsible use and a conduct rule, without a governance program behind them. A 2025 CARET blog post says the company regularly reviews how it uses or plans to use AI and the tools that support it. The Terms of Service bar customers from using the AI Services to generate unlawful, harmful or discriminatory content and ask them to report harmful outputs. Those are a statement of intent and a user obligation.

Nothing says how CARET governs its own models. No one is named as accountable, no framework such as ISO 42001 is cited, and nothing describes testing before a change to Quick Summary, deadline calculation or unbilled time detection reaches firms. No finding addresses whether those features perform unevenly across practice areas or matter types. Section 20.4 licenses CARET to train and tune models on firms' AI inputs and outputs, so firm data can shape future model behavior, and nothing describes the controls on that training.

The data processing addendum and security page cover most of the ground, without a published subprocessor list. The addendum, incorporated into the Terms of Service, commits CARET to report a personal data breach without undue delay and within 72 hours, investigate and mitigate, and help with the firm's own notices. It gives the firm 14 days after notice to object to a new subprocessor, audit rights that a third party audit report can satisfy, and a right to request audits of subprocessors.

Section 7 of the Terms permanently deletes client data from all CARET storage 30 days after the term ends. The security page describes daily backups with automatic failover, encryption of documents and data, TLS in transit, brute force detection, two factor authentication for external users, detailed logging, manual penetration testing and daily automated vulnerability scans by outside firms. No subprocessor list is published beyond Azure OpenAI, named on the blog. The addendum still cites the defunct Swiss U.S. Privacy Shield, and section 27 lets data transfer in a sale.

The allocation of risk is published in full, and on wrong AI output it stands behind nothing. Section 25 of the Terms of Service gives the firm a defense and indemnity for third party claims that the subscription service infringes intellectual property, with the usual remedies to procure a license, modify or terminate. Section 24 caps CARET's total liability at the subscription fees paid in the six months before the claim, a short look back, and excludes indirect and consequential loss.

Section 23 warrants only the 99.9 percent availability commitment, which excludes the AI Services, with a service credit as the sole remedy. It disclaims every warranty on AI outputs and says CARET is not liable for missed deadlines, late filings, sanctions, adverse rulings or malpractice claims arising from reliance on the service. Section 30 bars claims after two years, and the firm indemnifies CARET for its AI inputs. A firm can read exactly where it stands, but there is no remedy for a wrong summary or deadline.

Integrations are named, most with a line on what moves and which way. The integrations page lists two way email and calendar sync with Microsoft 365 and Google Workspace, IMAP for other mail, an Outlook plugin that tracks time while drafting and saves emails and attachments to a matter, and a Word plugin that opens and saves documents and versions directly to matter folders. NetDocuments can be reached from any matter, with a workspace created when a matter opens.

OneDrive folders can be linked to a matter to view, download, rename or delete files, and QuickBooks Online receives invoice approvals, edits and deletions automatically. HotDocs, RMail for encrypted email and electronic signature, LawToolBox for rules based deadline calculation and Lawmatics for lead and contact sync are named. Card, ACH and electronic check payments run in the product and the client portal. The Terms contemplate an API subject to rate limits. Setup documentation and field level detail for the connections are missing.

Cloud delivery is stated, and neither the hosting provider nor the region is. The Terms of Service define CARET Legal as a software as a service subscription, and the security page describes redundant servers with daily backups and automatic failover without naming who runs them or where. The 2023 and 2025 blog posts say Quick Summary runs on Microsoft's Azure OpenAI Service in a closed environment so client data is not exposed outside the firm's instance of CARET Legal, which implies separation between firms for the AI feature without saying whether the underlying infrastructure is shared.

The data processing addendum addresses transfers out of the European Economic Area, which tells a firm that its data may move across borders, but no region, residency option or processing location is stated. CARET still sells an on premise product, AbacusLaw, and one customer story describes moving from it to the cloud, but no private or single tenant deployment of CARET Legal itself is offered.

A named attestation and a payment security standard are stated, without a type, a date or a report that can be obtained. The security page says CARET Legal has achieved SOC 2 compliance and that payment card data is handled to the PCI DSS standard. It describes the testing behind that in detail: manual penetration testing of the application and infrastructure by its own security team, plus automated vulnerability scans by outside firms every 24 hours, along with encryption, TLS in transit, brute force protection, two factor authentication and activity logging.

The data processing addendum lets a third party audit report stand in for a customer audit, which implies a report exists. Nothing says whether the SOC 2 report is Type 1 or Type 2, its date, auditor or scope, including whether the AI features are covered, or how to request it. There is no trust center.

The provider behind the first AI feature is named, and the rest of the supply chain is not. CARET's 2023 launch announcement and a 2025 blog post say Quick Summary runs on Microsoft's Azure OpenAI Service in a closed environment, which names the model provider and hosting arrangement and tells a firm that Microsoft's enterprise terms sit underneath. The Terms of Service say the AI Services may use third party and publicly available models and that CARET's providers keep their rights in the underlying models.

Nothing says which models and versions run today, which providers stand behind the newer features for unbilled time, attention items and deadline calculation, where inference runs, or whether firms will be told before a model or provider changes. The data processing addendum gives firms 14 days to object after notice of a new subprocessor, but no subprocessor list is published to show whether AI providers are on it.

A figure, a unit and a term are published. CARET's pricing page lists three tiers billed annually per user per month: Enterprise at $79, Enterprise Plus at $99 and Enterprise Insights at $119, each with features listed, plus one time implementation fees. AI document summaries start on Enterprise Plus, and premium support costs extra, with a 25 user minimum. A firm can work out its annual cost for a given headcount before speaking to anyone.

The Terms of Service fill in the rest. AI Services may carry extra fees and token or credit limits set in the order form, and CARET may raise fees annually or on 30 days' notice for costs including cloud infrastructure and AI. Payments are not refundable, and ending before the term closes incurs the remaining balance. The pricing page's estimate that firms can capture an extra $20,000 of billable work per timekeeper is footnoted as an informational estimate. The cost of the AI token or credit limits is set in the order form and not published.

Segment and practice coverage are described with substance, and some limits are stated. CARET publishes solution pages for civil litigation, personal injury, estate planning, intellectual property, criminal, immigration, family, real estate, employment, insurance defense, workers' compensation and general practice. It also has role pages for firm administrators, partners, bookkeepers and IT, and pages for firms of one to thirty employees and more than thirty.

Customer stories cover an IP boutique in San Francisco, a real estate and tax firm in Kansas, a three office general practice in Ohio and an estate planning firm, and CARET's home page says more than 6,500 firms use it. Stated limits: the AI features begin on the second pricing tier, the service is governed by US law and subject to export control, and the Terms make firms responsible for confirming compliance with court rules in each jurisdiction.

Nothing says where the AI is weaker, for instance whether deadline calculation covers every state's court rules or relies on the separate LawToolBox integration.