Safety

AI assists. Clinicians decide.

Kadara is built so records, messages, and workflow outputs are reviewed by clinicians, patients consent to what is captured, and every action can be traced.

See how the safety model works ↓

How safety works in practice.

Kadara is designed so clinicians stay in control, patients consent to what is captured, and providers can review how actions happened.

  • AI assists, clinicians decide.

    Every Kadara output is a draft. The clinician reviews and approves before anything enters the record.

  • Every note requires clinician approval.

    Nothing becomes part of the clinical record until a clinician signs off.

  • Patient messaging is consent-based.

    Patients opt in before receiving summaries or reminders. They choose what they get and can stop at any time.

  • Approvals are traceable.

    Every view, edit, and approval is logged with user, timestamp, and device. The record is auditable end to end.

  • The system supports judgment, not replacement.

    Kadara is built to handle documentation load so clinicians can focus on care. It does not make clinical decisions.

Confidence scoring

Confidence scores guide the review.

Each SOAP section carries a confidence score that tells the clinician how certain Kadara is about the content, so review attention goes where it is needed.

Level Score What it means
Green 90% and above High confidence. One-click approval in most cases.
Amber 70% to 89% Moderate confidence. Clinician reviews this section before approving.
Red Below 70% Low confidence. Section needs rewriting or significant edits.
Audit and retention

Every action is logged.

Every view, edit, approval, and export is recorded with timestamp, user, and device. Audit logs are designed to be retained for 10 years in alignment with NIS ISO 27789:2019.

Data protection

Healthcare-grade data protection.

Kadara is designed to align with the Nigeria Data Protection Act 2023 (NDPA) and the General Application and Implementation Directive (GAID) 2025.

Encryption at rest (AES-256) and in transit (TLS 1.2 or higher).

Role-based access control with least privilege.

Auto-logout after 5 minutes of inactivity.

BYOD guardrails with biometric or PIN authentication.

Data hosted in Nigeria (AWS), with on-premise hosting designed for public hospital deployments.

72-hour breach notification to NDPC.

Patient consent

Patients control their data.

Patient summaries are strictly opt-in. Patients consent explicitly, choose what they receive, and can reply STOP at any time. Consent actions are logged with the rest of the record.

Clinical review

Ongoing clinical oversight.

Kadara is designed to be reviewed by a clinical advisory board across AI performance, bias detection, and patient safety. We commit to being transparent about what the system can and cannot do. Read the full clinical review model →

Clinical disclaimer

Kadara is an assistive tool, not a medical device. It does not provide diagnoses, treatment recommendations, or clinical advice. All clinical decisions remain with the licensed healthcare provider. AI-generated content may contain errors. Healthcare providers must review and edit before approval.

Want to go deeper on how this works?

Pre-launch, the safety approach is best discussed directly. We are happy to walk through how Kadara is designed to handle review, consent, and audit in your setting.

Speak with a specialist