Security for legal diligence

Core application compute is hosted in the EU. Database, authentication, document storage and vector search services are hosted in the United Kingdom. AI model providers are used through their global API services, with regions set by each provider.

• Core application compute: EU (EEA)
• Provider details available on request, under NDA where appropriate
• Customer legal practice is controller for Client and Matter Data
• YR Legal Services Ltd acts as processor for Client and Matter Data

Lexona uses AI models through provider APIs. Where configured and contractually available, model providers are used with enterprise privacy controls designed to limit provider retention and model training. Retention and training terms depend on each provider's agreement and account configuration.

• Lexona does not use Client or Matter Data to train AI models
• API-based model integration
• Model provider details available on request, under NDA where appropriate

Draft processor terms, data-handling responsibilities, and transfer materials are available for procurement and legal review. Each document's status is stated.

• Draft Article 28 Data Processing Agreement available for legal review
• DPIA support material available for practice review
• Data-subject request assistance set out in the draft DPA
• Draft DPA uses a without-undue-delay breach notification standard

Role-based and practice-scoped access controls are used. Access is restricted through authenticated service boundaries, and automated cross-practice access tests run in continuous integration on application changes.

• Practice scoping at the API and data-access layers
• Practice-specific vector search collections
• Multi-factor authentication is available to every platform user. A practice administrator can require it for everyone in their practice, and it is required for Lexona's own administrative accounts.
• Enterprise single sign-on (SAML/OIDC) on the roadmap

Audit logging and practitioner approval controls are used for applicable workflows. Registered agent runs create audit records, and supported workflows retain source, actor, timestamp and review events. Coverage varies by workflow and is documented during diligence.

• Audit records for registered agent runs
• Source provenance in supported conveyancing workflows
• Approval required before an AI-generated artifact can be exported
• Workflow-specific audit and export coverage documented during diligence

Practice-management integrations, starting with the proposed LEAP PMS integration, are designed around the principles below. Third-party integrations are subject to provider approval, API availability, applicable permissions and the terms of the relevant platform.

• Authorised users only, with permission-controlled, least-privilege access
• Only actions the platform's API supports, confirmed by the user
• Platform data read when a user runs a workflow, without syncing or keeping copies
• Auditable actions and minimised data
• Connected third-party platforms remain the system of record for data and workflows held within those platforms.
• No autonomous legal judgement and no unsupported changes to matter data