Security & Data Residency
Your residents' data deserves the highest level of protection. CaaS is built with security at every layer.
Australian Data Storage
All customer data, including voice recordings, transcripts, and compliance evidence, is stored in AWS Sydney (Australia). AI processing (Anthropic for language; LiveKit, Deepgram and ElevenLabs for voice) currently occurs in the United States under Data Processing Agreements. Anthropic's enterprise terms prohibit training on customer data by default, and equivalent written confirmations from the voice vendors form part of our supplier review.
End-to-End Encryption
AES-256 encryption for all data at rest. TLS 1.3 for all data in transit. Encryption keys are managed using industry-standard key management services with automatic rotation.
Role-Based Access Control (RBAC)
Granular access control ensures staff only see data relevant to their role and facility. Coordinators see their unit. Managers see their facility. Admins see their organisation. No more, no less.
Row-Level Security (RLS)
Database-level security policies enforce access control at the row level. Even if application logic fails, the database itself prevents unauthorised data access. Defence in depth.
Comprehensive Audit Logging
Every access, modification, and export is logged with timestamps, user identity, and action details. Audit logs are designed to be tamper-resistant and retained for the applicable regulatory period.
Full Security Documentation Portal
Access our comprehensive security documentation, policies, and compliance reports through our dedicated security portal.
Visit security.careplans.ioISO 27001 ISMS
careplans ai maintains an Information Security Management System (ISMS) aligned to the ISO 27001:2022 standard. This framework governs how we identify, assess, and manage information security risks across all aspects of the platform.
Trailing Stop AI Safety
The Trailing Stop is an independent AI safety system that monitors every call in real time. It operates as a separate LLM instance, independent of the conversation model, continuously evaluating the interaction for safety concerns.
How It Works
A secondary LLM processes the conversation stream in parallel, evaluating for distress signals, clinical risk indicators, and situations requiring human intervention. If risk thresholds are breached, it triggers automatic escalation to on-call staff, can adjust the conversation tone, or gracefully end the call with appropriate care messaging.
Defence in Depth
The Trailing Stop cannot be influenced or overridden by the conversation model. Its safety decisions are final and logged immutably. This creates a true defence-in-depth architecture for AI safety in clinical settings.
Supplier Certifications
| Supplier | Service | Certifications |
|---|---|---|
| AWS (Sydney Region) | Cloud Infrastructure | SOC 2 Type IIISO 27001IRAP ProtectedPCI DSS |
| Supabase | Database & Auth | SOC 2 Type IIISO 27001HIPAA |
| Anthropic | AI / LLM Provider | SOC 2 Type IIISO 27001 |
| Twilio | Voice / Telephony | SOC 2 Type IIISO 27001PCI DSSHIPAA |
Need More Detail?
Request access to our full security assessment documentation, including control self-assessments and compliance documentation.
Request Security Assessment Access