Product and intended use
Define the configuration, people, information and tasks being assessed.
Cloud, private cloud, on-prem
Public claim only
HAQQ Inc. Terms 29 June 2026. Security: HAQQ Cloud in a chosen region; Private Cloud in the customer's cloud tenancy with customer-held keys; On-Premise on customer hardware. Product modules include e-Firm, Legal AI (Justinian), e-Client. Credits meter Legal AI.
Ask: Which hosting model, region and modules are on our Order Form?
EU, US or Middle East
Public claim only
Security: available regions are the European Union, the United States and the Middle East. Privacy N: in-region hosting by default where committed in an Order Form or configuration; cross-border transfers may still occur for support, security, billing, subprocessors, redundancy or law.
Ask: Write the region into the Order Form and list any permitted support/subprocessor exceptions.
Legal AI agents and e-Firm
Public claim only
Terms define Legal AI (retrieval, summarisation, drafting, extraction, workflow assistance, AI Agent, Firm Digital Twin) and e-Firm matter management. Security: seven-layer Justinian architecture with guardrails and configurable human approval for consequential agent actions.
Ask: Which Legal AI features and agent tools are enabled on our tenant?
No Australian corpus described
Not confirmed
Privacy lists GDPR, UK GDPR and several Middle Eastern privacy laws as examples. No licensed Australian case-law or legislation corpus was described on pages reviewed.
Ask: Does Legal AI retrieve Australian primary law, or only customer-uploaded and other named sources?
Data, controls and requirements
Compare the actual configuration with firm, client and jurisdiction requirements, supported by evidence.
Customer warrants lawful upload
Public claim only
Privacy J: Customer must have notices, consents, professional authorisations and a lawful basis for data it submits, including sensitive matter files. No Australian client-disclosure template was published.
Ask: Provide a client-disclosure for EU/US/ME hosting and Legal AI, which we can adapt for APP 8.
Providers not named publicly
Not confirmed
Security: foundation-model layer with no training and BYOK. Privacy authorises AI infrastructure subprocessors but does not name OpenAI/Anthropic/Google on the pages reviewed. Model mix is therefore unknown.
Ask: Name every model provider, region and whether ZDR applies on our tenant.
List on request; 30-day notice
Public claim only
Privacy L: customer authorises subprocessors; HAQQ will maintain a list (platform, Trust Center, email, schedule or on request) with function and location where practicable. 30 days' prior written notice of new material subprocessors (shorter for emergencies); 15 days to object, then alternatives or terminate the affected Product.
Ask: Provide the current named list with countries before we upload matter data.
No training except signed agreement
Public claim only
Privacy B and DPA summary: HAQQ's internal product development does not involve training, fine-tuning or improving AI models on Customer Data, Matter Data, prompts, outputs or User content, except where expressly agreed in a separately signed written agreement and permitted by law. Security FAQ: never used to train.
Ask: Confirm the Order Form contains no training carve-out and that subprocessors are on ZDR/no-train terms.
Anonymised telemetry
Public claim only
Privacy: Derived Data is anonymised/aggregated telemetry for performance, reliability, capacity, diagnostics and security, and must not identify Customers, Users, Clients or data subjects. It may be retained indefinitely once irreversible.
Ask: Can prompt text appear in telemetry before anonymisation, and can we disable product analytics?
Customer-configured plus 90-day exit window
Public claim only
Security FAQ: customer configures retention; purging is automatic and produces a verification log; export anytime. Privacy Q: 90-day post-term export window, then delete from active production within a commercially reasonable period; backups up to one year.
Ask: Set our retention to 30 days or less for prompts and matter files, and confirm backup TTL.
Return or delete; certification on request
Public claim only
Privacy Q: customer elects return or delete in writing; default is delete if no election. Active-system deletion after the 90-day window; backups then anonymised or deleted when feasible. Written confirmation of deletion may be provided on request, subject to limitations.
Ask: Will you certify deletion of production, backups and subprocessors within a stated number of days?
Nobody by default; support needs approval
Public claim only
Security FAQ: nobody at HAQQ can see customer data by default; access is role-based, least-privilege and logged; support access to a specific matter requires explicit approval and expires. Privacy I: authorised persons bound by confidentiality.
Ask: Is support access ticket-scoped in the product, and can we refuse it?
SSO, MFA, RBAC, ethical walls
Public claim only
Security: SSO, MFA, role-based access; one tenant per firm; no cross-matter mixing; ethical walls at the governance layer. Privacy: Customer configures permissions and retention.
Ask: Will SSO+MFA be mandatory, and how are ethical walls tested?
TLS 1.3 and BYOK claimed
Public claim only
Security: encryption in transit including TLS 1.3 where supported; enterprise customers can bring their own key; revoking the key ends HAQQ's ability to read the data. Privacy M lists encryption in transit and at rest and customer-managed keys where available for applicable tiers.
Ask: Is BYOK in our SKU, and what at-rest algorithm is used on HAQQ Cloud?
Alignment, not completed certification
Not confirmed
Security header lists SOC 2, ISO 27001, ISO 42001 and GDPR. FAQ: 'Compliant. ... We describe that as alignment rather than a completed third-party certification.' Privacy M: framework references are not a representation that every product is in the same certification scope. Treat as uncertified unless a report is produced.
Ask: Produce the current SOC 2 / ISO certificates with dates and scope, or confirm none exist yet.
Published Book III DPA; Delaware terms
Public claim only
Book III of the Terms operates as the DPA (privacy notice hosts the full text). Accepting Terms puts it in force; a counter-signed copy is available on request. Delaware law/courts. Audit rights in Privacy O (summaries first; on-site with 30 days' notice, once per 12 months).
Ask: Will you counter-sign Book III plus SCCs and our APP 8 addendum?
Region on the Order Form; SCCs available
Public claim only
Residency where committed. Privacy N: transfers may use adequacy, SCCs, UK addenda, supplementary measures or other lawful mechanisms. Cross-border still possible for support, billing, subprocessors and redundancy. Australia is not a listed region.
Ask: Map every country of storage, inference and support for our Order Form.
72-hour outer target
Public claim only
Privacy P: notify without undue delay after becoming aware; commercially reasonable efforts as soon as practicable and, where feasible, within 72 hours — described as a contractual outer target, not a licence to delay. Security FAQ also states 72 hours per GDPR. Customer remains responsible for notifying regulators/clients unless HAQQ has a direct duty.
Ask: Name the notification email path and confirm 72 hours applies to our Cloud tenant.
30-day subprocessor notice
Public claim only
New material subprocessors: 30 days' written notice with function and location; 15 days to object. Terms hierarchy: signed agreement, then Official Proposal/Order Form, then Book III for privacy. Product Books can impose more specific obligations.
Ask: Will material model or region changes also get 30 days' written notice?
90-day export then delete
Public claim only
Privacy Q: 90-day export window in a structured, commonly used, machine-readable format; then delete (or return if elected). Backups up to one year. Certification on written request. Security FAQ: export anytime while live.
Ask: What export formats are supported, and will backups be in the deletion certificate?
Integrations layer not named
Not confirmed
Architecture includes a data-and-integrations layer. Named third-party connectors were not listed on the security or privacy pages reviewed. AI Agents may invoke permitted tools within tenant permissions.
Ask: List every connector and tool the agent may call on our tenant, and how we disable them.
Not described
Not confirmed
Legal AI is described as retrieval over customer-authorised sources. Open-web search was not stated on pages reviewed.
Ask: Can Legal AI browse the open web, and can we disable it?
Configurable human approval
Public claim only
Security: Justinian agentic harness has guardrails, audit trails and configurable human approval for consequential agent actions. Terms: agents plan and execute steps including retrieving, drafting or invoking permitted tools within customer-configured permissions.
Ask: Which actions are 'consequential' by default, and can we require approval for every external call?
Agent does not exercise legal judgment
Public claim only
Terms: an AI Agent does not exercise independent legal judgment. Every action is claimed to be logged, attributable and exportable. No in-product mandatory sign-off before a lawyer relies on output was extracted.
Ask: Is there a mandatory human-accept step before exporting a draft or sending e-Client messages?