The person on the other end of that call is someone's whole world. We treat them that way.

KindredMind is designed around privacy, human oversight, clear limits, and respect for the person using the service. A family may configure an account, but that practical role is not automatically the same as voice-owner consent, the interacting person’s willingness, or formal substitute decision-making authority.

Our safeguards support thoughtful family decisions; they do not declare that one arrangement is right for every person.

Data minimization

Families do not need to share medical records, diagnoses, or unnecessary sensitive details. We do not ask for a loved one’s full legal name or date of birth. Share only the context needed for the chosen support, review who can access it, and remove it when it is no longer needed.

Consent, participation, and authority

There are several separate questions when KindredMind is introduced:

• Account authorization: who creates and manages the account and settings.
• Voice-owner consent: the person whose voice is used should provide appropriately documented consent for that use.
• User willingness and assent: the person interacting with the system should be involved to the extent appropriate to their abilities, preferences, and circumstances.
• Substitute decision-making: where assistance is actually required, use an appropriately authorized decision-maker where applicable for that specific decision.

A dementia diagnosis does not automatically establish incapacity. Ability can vary between people, by decision, and over time. A spouse, adult child, caregiver, or account holder does not automatically have legal authority to make every decision for another adult. This page does not make legal determinations; when authority is uncertain, seek advice appropriate to the relevant jurisdiction.

KindredMind can be configured by a family member or caregiver. That account role does not erase the interacting person’s preferences. Explain the service in an accessible way, invite questions, observe willingness, and pause or stop if the person does not want to continue.

Transparency and AI identity

KindredMind does not rely on a universal policy that conceals AI identity from people with dementia, and we do not claim that dementia-care ethics proves nondisclosure is always right. There is no single answer for every person or situation. Proportionate transparency should take account of what the person understands, their preferences, the nature of the interaction, the risk of material deception, and any applicable authority.

Someone does not need to understand every technical detail before any assistive technology can be considered. They do deserve dignity, a meaningful opportunity to participate, and a response to questions that does not intentionally create a materially false belief. If the interaction causes confusion, distress, or unwanted intrusion, stop and reconsider. Emotional wellbeing matters, but it does not broadly override factual transparency in every situation.

Voice setup and family context

The voice owner’s interview-session recordings are used only to build that voice model and are automatically deleted within 24 hours of the model being ready. Real conversations are not used to train models. If the voice owner does not want to record, an assistant voice may be available.

The account holder adds the family-provided context used for supported conversation. The voice owner, account holder, and user can be different people. Consent for one role should not be treated as consent for all roles.

Privacy, encryption, and deletion

KindredMind collects account information, family-provided context, voice-model source recordings, and information generated by supported calls. Sensitive data is encrypted at rest with AES-256; data in transit uses TLS 1.3; and voice presence files are stored in encrypted object storage with restricted access.

• Call transcripts are encrypted and automatically deleted after 90 days.
• Family conversations and intake information are not used to train AI models.
• Families can export available data from account settings and request deletion.
• The voice model can be deleted from the dashboard and is permanently removed within thirty days of a request.

Safety and human oversight

Configured alerts and post-call summaries can help a caregiver review supported calls, but KindredMind is not a medical monitoring device, emergency service, or guarantee that every concern will be detected. Keep a human route for urgent help. Do not rely on the system for diagnosis, emergency response, or a decision that requires professional judgment.

The service should not promise that a caregiver is on the way when they are not, and it should not replace visits, human calls, clinicians, staff, or emergency services. Review whether the product remains wanted and appropriate for this individual.

Your control and exit

The account holder can configure supported call handling, review available summaries and transcripts, and turn the service off. Agree in advance who maintains settings, who can access records, and how the person using the system can ask for a pause. A request to stop is a valid outcome.

Frequently asked questions

Further reading

For the broader ethics framework, read Is AI ethical in dementia care?. For practical technology evaluation, read How to Evaluate Dementia Care Technology.