One distinction needs to be fixed before the market fixes it badly.
A future AI system may look like a long-lived digital companion, assistant, twin, or “entity”. It may remember, speak consistently, keep a durable interface, and appear continuous over time.
But surface behaviour does not answer the real governance question:
Who controls the root of its identity chain?
That is the problem behind my new technical note:
Entity vs. Profile: A Witness-Root Custody Criterion for Persistent Digital Entities v0.1.1
The note proposes a narrow custodial criterion.
Do not classify by behaviour. Do not classify by branding. Do not classify by model capability. Classify by witness-root custody.
A persistent digital construction may be classified as an entity only when its witness root is generated and held inside its closed-box boundary under a-side lineage custody.
If the witness root, usable copy, escrow, registry authority, platform root, recovery authority, or override capability is held outside that boundary, the construction is a profile — even if it behaves like an entity.
Profiles are legitimate products. There will be useful digital profiles, companions, personal assistants, memory systems, and AI twins.
The point is not to prohibit them.
The point is to avoid a standards-level conflation where externally rooted persistent digital profiles absorb the stronger term “entity” before the distinction is named.
Because once the language is captured, later correction becomes much harder.
This is not a personhood claim. Not certification. Not AI Act compliance. Not a conformity assessment. Not a deployment authorization.
It is a terminology and evidence-model proposal.
Maximum claim strength: EVP-C3 — evidence model proposal.
Reader page: https://ivankotov.eu/publications/entity-vs-profile-v0-1-1/
Version DOI: https://zenodo.org/records/21204415
Zenodo: https://zenodo.org/records/21204415
GitHub metadata bridge: https://github.com/Kot141078/advanced-global-intelligence/tree/main/official/entity-vs-profile/v0_1_1