- Status
- Published
- Version
- 1.0
- Effective
- 4 August 2026
- Last reviewed
- 4 August 2026
- Review cycle
- Annually
- Last updated
- 4 August 2026
Purpose
To define how artificial-intelligence assistance may be used within Foundation and partner platforms, the boundaries that apply, and the safety commitments that protect people using it.
Scope
Any AI-assisted feature offered through Foundation-operated services or through the member workspace operated with our partner platform Funeora.
Responsibilities
Platform administrators are responsible for configuration, monitoring and escalation routing. Qualified professionals remain responsible for clinical and specialised care.
What AI may be used for
- Helping a person put words to what they are experiencing.
- Reflective prompts, journalling support and general emotional-wellbeing information.
- Signposting to Foundation programmes, community services and professional pathways.
- Administrative assistance such as drafting, summarising or organising a member's own content at their request.
What AI is never used for
- Diagnosis, psychological assessment, psychiatric evaluation or prescription.
- Crisis management or emergency response.
- Replacing a psychologist, counsellor, psychiatrist, social worker, doctor or safeguarding officer.
- Making decisions about a person's care, eligibility, employment or support without human review.
- Training external models on member content.
Disclosure
Where a response is generated with AI assistance, this is disclosed clearly. AI is never presented as a human, a clinician or a Foundation staff member.
Safety and escalation
Where a conversation indicates risk of harm to the person or to someone else, the feature surfaces human and emergency pathways immediately and encourages contact with an appropriate professional or emergency service. Automated support is not treated as a sufficient response to risk.
In an emergency in South Africa, contact emergency services on 10111, or 112 from a mobile phone.
Data handling
- Member content is not shared with AI services by default; assistance runs only on the content the member submits for that purpose.
- Content submitted to an AI feature is not used to train external models.
- Safeguarding, health and identifiable beneficiary records are excluded from AI processing.
- Retention of AI interaction records follows our Data Retention & Deletion Policy.
Limits and review
AI responses can be incomplete or wrong. Members are asked to treat them as general information, not professional advice. AI features are reviewed regularly, and any feature that cannot be operated safely is withdrawn.
Questions about this policy? Email info@eflegacy.org and a member of our team will respond within five working days.