Product and intended use
Define the configuration, people, information and tasks being assessed.
Product and plan covered
Public claim only
AU Clio is sold as Manage (EasyStart–Elite), Work (standalone or Signature/Elite Work packages), Grow (intake/CRM) and Draft (document automation). Manage AI is included on Core and above, not EasyStart. Grow AI is usage-priced per hired matter. Images on the AU legal-AI page still use the North American 'Clio Duo' filename; current AU product names are Manage AI and Work.
Ask: Confirm our Australian order form: Manage tier, whether Work/Grow/Draft are included, Grow AI and Docket metered rates, and whether any vLex Vincent SKU is separate.
Deployment model
Public claim only
Clio is multi-tenant SaaS. The security page states hosting choices in Canada, the US, Europe and Australia on AWS, with geo-redundant backups and optional data escrow. No public on-premises option was found.
Ask: Is our tenant exclusively in the Australian AWS region, including replicas, logs, Work indexes and Clio Operate logging?
AI features in scope
Public claim only
Manage AI (Core+) drafts bills, turns documents into calendar events and tasks, and drafts client updates from matter activity. Work analyses matter files, researches Clio Library, drafts, supports Clio for Word, optional Web Search and Clio Docket, and is described as agentic. Grow AI offers chat, email and (where available) voice agents. Draft adds templates, auto-fill and generative drafting.
Ask: Which of Manage AI, Work (including Web Search, Docket and Word), Grow AI channels, and Draft will be enabled for which roles?
Jurisdictions and legal content
Public claim only
AU terms are governed by New South Wales law with exclusive Sydney courts. Work is marketed with Clio Library (100+ countries) and Cert citator; the AU launch (18 August 2026) says it is built on verifiable Australian legal data. State-by-state collections and currency rules are not published on the pages read.
Ask: List the Australian courts, legislation and citator coverage Work actually searches for our practice areas, and how unreported or very recent authorities are treated.
Data, controls and requirements
Compare the actual configuration with firm, client and jurisdiction requirements, supported by evidence.
Client authority and professional use
Not confirmed
AU terms put client notices and consents on the firm (section 2.17) and make the firm responsible for end-user AI disclosures (section 14.4). Grow AI is designed to tell prospects they are interacting with AI; voice calls play a non-disableable recording notice. No template court-practice-note wording was found. Putting matter files into Work or Manage AI is the firm's authority decision.
Ask: What client-disclosure wording, matter-exclusion controls and court-practice-note records (for example NSW SC Gen 23) does Clio supply, and can we block AI on nominated matters?
Foundation-model providers
Public claim only
The AU subprocessor list (modified 20 August 2026) names AWS Bedrock, Anthropic, OpenAI, Google Vertex AI and Microsoft Azure Foundry as AI processors for Manage, Grow, Draft, Work, Vincent AI and (in part) Clio Operate. Locations include the United States and EU; Vertex also lists Canada and Australia. Anthropic is listed as United States only. There is no customer-facing model picker on the pages read.
Ask: For our Australian tenant, name the live model, legal entity and endpoint region for Manage AI, Work, Grow AI and Draft, and the notice period before a provider is swapped.
Subprocessors
Public claim only
AU Exhibit C (20 August 2026) lists infrastructure (AWS, Azure for Operate, Google for Docket Alarm), AI processors (including Braintrust, Turbopuffer, Vapi for Grow), feature providers, payment processors, and affiliates (including vLex entities, Fastcase and Themis Solutions Inc. in Canada). The DPA lets Themis use listed subprocessors and notify of new ones; a reasonable objection can lead to moving data or 30 days' termination. Notice is 'from time to time', not a fixed period.
Ask: Confirm the current dated register for our AU tenant, including Braintrust and Vapi if Grow AI is on, and the written notice we will get before a new AI subprocessor is added.
Training on customer content
Not confirmed
AU terms section 14.6 say Themis may de-identify and aggregate Content submitted to, and Output from, AI Services to improve and ensure the quality of the AI Services, and that use of AI Services does not grant Themis the right to train generalized LLMs on Confidential Information. Pricing, Work, legal-AI and security pages instead say firm data is never used to train AI models or for any other external purpose, and that ZDR agreements mean providers do not store or train on the data. Those marketing lines are broader than clause 14.6.
Ask: Which signed document prevails, and does it prohibit training and product improvement on prompts, outputs, uploads, embeddings, feedback and telemetry by Clio, OpenAI, Anthropic, Google, Amazon, Microsoft and Braintrust?
Feedback and quality monitoring
Public claim only
Clause 14.6 allows de-identified aggregated AI Content and Output to improve AI quality. Braintrust Data, Inc. is listed as 'AI Analytics, Evaluation and Testing' (United States) for Manage, Grow, Draft, Work and Vincent AI. ProductBoard and Qualtrics are listed as Themis (controller) providers for feedback and insights, not as processor subprocessors.
Ask: Can we disable any in-product rating or Braintrust evaluation on our tenant, and is feedback content excluded from model training and from access by staff outside a ticketed review?
Retention of prompts, outputs and matter indexes
Public claim only
Work and package FAQs claim zero retention at third-party AI providers. Platform Content is retained for the subscription and 90 days after cancellation. Privacy retention is purpose-based. Vector database Turbopuffer is listed in the US, EU, Canada and Australia. No published default for Work chat history or embeddings after a matter is closed.
Ask: State default and configurable retention for Manage AI and Work chats, embeddings, Grow transcripts/recordings, Braintrust copies, logs and backups for the Australian tenant.
Deletion and certified erasure
Public claim only
After cancellation, Content is available for at least 90 days, then 'irrevocably deleted'. Escrowed data remains six months. Privacy offers deletion via account or a Securiti DSAR form, with a 30-day response. No public certified-deletion SLA covering AI subprocessors was found.
Ask: On exit or matter close, what is deleted (chats, embeddings, Grow recordings, Braintrust, subprocessors), in what period, and will you certify deletion in writing?
Vendor and verifier access to content
Public claim only
AU terms section 21.3 say only Themis, with strict business reasons, may access and transfer Content to provide the Service, with reasonable prior notice. Affiliates (Canada, US, UK, Ireland, Spain, vLex entities and others) are subprocessors who may have access. Gong, Talkdesk and Zoom are listed for call recording on the Themis (controller) list. No lawyer-verifier network like LawY was described for Work.
Ask: Who can read our matter content (Themis AU/CA/US, vLex, model providers, Braintrust), under what ticketed approval, from which countries, and can we see those access logs?
Access control, SSO and audit
Public claim only
Administrators manage users. Help centre documents SAML SSO from Manage or Grow (Okta, Azure, or custom IdP), with a recommended break-glass admin exclusion. Security page offers SAML SSO and MFA. Twilio is listed for SMS MFA. SCIM provisioning was not described on the pages read. Legal-AI copy claims permission awareness and auditability without showing an AI audit-log export.
Ask: Confirm SSO/SCIM and mandatory MFA on our plan, default-off AI for nominated users/matters, and whether we can export a complete Manage AI and Work audit log.
Encryption in transit and at rest
Public claim only
The security page states HTTPS/TLS 1.2 or higher in transit and AES-256 at rest. AU terms say Content is stored securely and encrypted. Customer-managed keys were not offered on the pages read.
Ask: Confirm encryption of Work indexes, Grow recordings, logs and the path to each model provider, and whether customer-managed keys are available for the Australian tenant.
Independent security assurance
Public claim only
Clio claims annual SOC 2 Type II and SOC 1 Type II examinations, with reports at trust.clio.com. AU terms section 5.4 offer a SOC 2 or SOC 3 (or successor) report for data-centre facilities within 30 days, subject to the auditor's agreement. ISO 27001 for Clio itself was not claimed on the pages read. Reports were not inspected; AI-feature scope is unverified.
Ask: Provide the current SOC 2 Type II report covering Manage AI, Work, Grow AI and Draft, with scope, exceptions, bridge letter and penetration-test summary.
Contract, DPA, liability and forum
Public claim only
AU ToS (4 August 2026) are NSW law, Sydney courts. Liability is heavily excluded except fraud, negligence or wilful misconduct, with ACL-style limitation to re-supply. AI Output is as-is; the firm indemnifies Themis for end-user AI deployment. A DPA is published (SCCs module 2; 72-hour incident notice; subprocessor objection). Standalone Draft has separate terms. vLex terms may apply to library content.
Ask: Confirm the DPA is incorporated for our AU order, whether vLex terms apply to Work research, and any negotiated override of the AI as-is clause and improvement licence in 14.6.
Cross-border transfers and APP 8
Not confirmed
Security says AI data is encrypted and processed in the customer's region (US, Canada, EMEA or APAC) and that hosting can be in Australia. The AU subprocessor list still places OpenAI, Anthropic, Bedrock and Azure Foundry in the United States and/or EU (Vertex also lists Australia). AU privacy says personal information of Australian customers may be processed outside Australia, including in the United States, under DPAs and similar safeguards.
Ask: For Manage AI, Work, Grow AI and Draft, list every country for storage, inference, backups, support and logs, the transfer mechanism, and whether Australia-only inference is actually configured for us.
Incident notification and history
Public claim only
AU terms section 5.5 require Themis to report a Security Breach within 72 hours of learning of it. The DPA repeats 72-hour Security Incident notification and assistance. Incident.io is listed as Themis's incident-management provider. No public incident history for Work or Manage AI was found.
Ask: Confirm 72 hours runs from discovery not public confirmation, covers AI subprocessors, includes NDB/OAIC cooperation, and disclose any material incidents in the last 24 months.
Change notice for models, terms and processors
Public claim only
The DPA notifies new subprocessors 'from time to time' with a reasonable-objection path. Grow AI usage fees may be updated without advance notice. Product-specific Web Search and Docket fees may change. No published freeze of the model list was found.
Ask: What written notice will we get before a new model, subprocessor or Web Search default is introduced, and can we object or freeze the model list?
Export and exit
Public claim only
On cancellation, the Administrator has 90 days to retrieve Content, then deletion. Optional data escrow lasts six months. Work can be cancelled effective end of the current monthly or annual term. Export formats for Work chats versus Manage matters are not specified. A portability blog exists on the AU site.
Ask: What export formats cover matters, documents, Manage AI and Work history, and will you certify deletion from Clio and AI subprocessors after the 90-day window?
Connectors and third-party AI tools
Not confirmed
AU terms treat Third Party Services as outside Clio warranties; Themis may disable integrations without notice. The app directory and Dropbox Sign, Nintex, Bandwidth, Stripe and Plaid appear on the subprocessor/feature lists. Enabling a connector is the firm's configuration choice.
Ask: List every connector that will be on for our tenant, the data each receives, and confirm they are out of scope unless we enable them in writing.
Open-web search
Public claim only
Product-specific terms section 2 describe optional Web Search: AI Services may query publicly available websites to supplement Materials. Output may include third-party site content that the firm must verify. Work marketing also says Clio Library is 'not scraped from the web'. Whether Web Search is on by default in AU Work is not stated.
Ask: Is Web Search on or off by default for our AU Work tenant, can we disable it firm-wide, and which sites may be queried?
Agent actions and confirmation
Not confirmed
Manage AI is described as turning court documents into calendar events, matter activity into client updates, and time into invoices ready for approval. The AU Work launch says agentic capabilities plan and execute multi-step tasks with visibility and control. Grow AI can chat, email and (where offered) answer voice calls. Product terms say Drafting Features include agentic read/analyse/edit. Confirmation gates are not fully specified per action.
Ask: Which Manage AI, Work and Grow AI actions can execute without a second confirmation, and can we disable calendar writes, client-message send and Grow voice?
Human review before use
Public claim only
AU terms section 14.2 require human oversight and review of Output for accuracy and professional duties before disclosure or use. Marketing repeats that the user reviews and approves before a client or the court. Grow AI disclosures to prospects are partly non-configurable. Nothing in the terms forces a second solicitor to approve a Manage AI client update before send.
Ask: What product controls force a named reviewer before client email, portal send, Grow AI replies, or saving a Work draft to the matter, and can those controls be mandated firm-wide?
