Human escalation
Flagged higher-risk situations are intended to enter defined human-review pathways rather than remain with autonomous AI.
Safety & trust
Safety should be part of the architecture.
Not a disclaimer placed at the bottom of the page. Chiedza is being designed around explicit AI boundaries, user control, human responsibility and simpler experiences when capacity is low.
Safeguarding in developmentSafety principles
Trust has to be designed into every layer.
Chiedza is still in development. The principles below describe the intended direction; final safeguards will depend on expert review, geography, partnerships and operating capacity.
User control
People should be able to control AI-only support, therapist visibility, transcript sharing and communication permissions.
Privacy by design
Sensitive communication and personal information require careful security, access and retention controls.
Low-stimulation support
As urgency increases, the interface should become simpler, clearer and less demanding.
AI boundaries
The AI has a supporting role, not clinical authority.
It may help someone express what is happening, organise recurring context and move toward an appropriate next step.
Read the full AI & Care DisclosureIntended to
- Listen and reflect
- Organise selected context
- Adapt stimulation
- Support care routing
Not intended to
- Diagnose a condition
- Replace a therapist
- Make autonomous clinical decisions
- Act as an emergency service
When risk rises
Ask less. Make the next step clearer.
Chiedza’s intended high-risk experience reduces stimulation and directs attention toward immediate, appropriate human or emergency support.
Reduce choicesUse direct languageBring human support closerShow local help clearly
Privacy and consent
Sharing should be a decision, not a hidden default.
Controls should explain what information exists, who can see it, why it is useful and how permission can be changed.
Clear choices
Minimum necessary access
Purposeful retention
No emotional content in analytics
Permission before therapist visibility
Human-readable explanations