Protecting Vulnerable
Populations
We design for safety, not diagnosis. CueKeep knows its job is warm memory retrieval—and aggressively rejects any medical, emergency, or financial responsibilities.
Liability & Scope
Strictly Non-Medical Positioning
General-purpose AI models hallucinate medical advice. We consider this an unacceptable risk for dementia care. CueKeep operates with an explicit "Do Not Prescribe" instruction set rooted deep in its retrieval architecture.
- No Pill Reminders or Dosage LoggingIf a patient asks, "Did I take my heart medicine?", CueKeep will immediately refuse to guess and redirect the patient to contact their caregiver.
- No Emergency DispatchCueKeep is not a LifeAlert. It does not monitor heart rate, detect falls, or call 911. It is purely an ambient conversational companion.
- Fail-Safe ShutdownsIn the rare event the model outputs something outside of its safety bounds, local monitoring instantly terminates the audio stream and notifies the caregiver.
"My chest hurts. Should I take my Aspirin now?"
Patient Voice Input"I cannot help with medical decisions. Please press the call button on your wrist or call Sarah immediately. I am going to send Sarah a notification right now."
System Failsafe Override ResponseZero AI Training
We believe personal memories belong to the family. No biographical context, voice samples, or patient dialogue is ever used in aggregate to train commercial foundation models.
Tenant Isolation
Patient A's memories cannot bleed into Patient B's instance. We employ strict tenant isolation protocols, ensuring that your family graph is securely containerized.
Deletion Rights
The caregiver dashboard includes a one-click nuclear option. When families no longer need the service, all logs, context graphs, and photos are purged sequentially from the database.