Risks Of AI Clones

The most concrete risk is straightforward: a clone of you, built without your permission, can be used to ask your family for money, your employer for access, or your bank for a transfer. Voice cloning alone has already enabled multi-million-dollar wire fraud.

The defense is partly technical (liveness checks, voice watermarking, callback verification) and partly social (households and companies that practice 'verify out of band' as a default).

Consent for an AI clone is not a single click. It is a long, mutable relationship: what the clone can say, to whom, where, and for how long. Most current product flows compress all of that into a single check box.

A meaningful consent regime treats the clone as an ongoing instrument of the person and gives them durable controls — including the right to suspend, retrain, or destroy the clone at any time.

A clone of a loved one is a powerful thing to interact with. For some people it eases grief; for others it deepens and prolongs it. Therapists are only beginning to develop guidance for clients who form daily relationships with the digital remains of someone they loved.

The risk is not unique to grief clones. Any clone that becomes a primary interlocutor — a parent, a mentor, a partner — competes with the relationships it pretends to support.

Most clones today live on hosted platforms whose terms of service allow far more than users realize: retention of voice prints after account deletion, secondary training rights, and surprise policy changes.

The structural risk is that the most personal possible technology ends up entirely owned by the companies that host it. Portability standards and explicit deletion guarantees are still missing from almost every major provider.