Three positions
Position 1
Strict disclosure
Every clone-generated interaction is labeled at the start and at any point the user could reasonably forget. No exceptions.
Position 2
Context-aware disclosure
Disclosure intensity scales with stakes. Casual scheduling can be one-time; medical advice or political speech requires repeated, persistent labels.
Position 3
Functional disclosure
Provenance metadata embedded in the artifact is sufficient. Human-visible labels create alarm fatigue and are ignored.
Best practices
- ●Stake-weighted labels. The higher the consequence, the more conspicuous the label.
- ●Provenance plus visible label, not one or the other.
- ●User-facing log. Recipients can see at any time which interactions were clone-mediated.
