Practical independence.
An independent personal system could recognise escalation, identify coercive patterns, preserve evidence privately, contact appropriate support, arrange autonomous transport, locate safe accommodation and organise food, medication, childcare and documents.
It could connect medical, legal and psychological assistance without asking a person already under pressure to navigate a fragmented system alone.
Design boundary: safety systems must be independent, privacy-preserving and controlled by the person at risk. AI itself is not automatically trustworthy.
Remove dependence
without removing agency.
The ambition is not surveillance or forced optimisation. It is to make safe alternatives available before cruelty becomes destiny.