AI identity
Documented architectureA stable agent identifier links a declared model or service context to its current policy and authority record.
Preview · not officially published. Technical information only; the planned DLT is not live.
Builders
Sig-56, Web4 and F.I.P. appear in the project's own material and are described here at the maturity they actually hold: concepts and positioning terms. None is a ratified standard, none is certified, and no specification document has been supplied for inspection.
The platform claims no conformance with, endorsement by, or membership of any standards organisation. Where an external specification is eventually referenced, it will be named precisely, with its version and its publisher.
Describing an idea in standards language does not make it a standard. These pages keep the two apart on purpose.
A signature and signalling concept for binding identity, authority, rule context and intent to a verifiable action, so that authorization remains reviewable across governance and execution boundaries.
| Element | Current content |
|---|---|
| Purpose | Make an authorized action checkable long after the fact, independently of the system that produced it. |
| Scope | Binding of identity, mandate, rule version and intent to an action record. |
| Actors | Subject, authorizing body, evaluating protocol, later verifier. |
| Data / proof objects | Signed action binding with authority and rule references. Structure to be specified. |
| Lifecycle | Intent → authority check → binding → acceptance → anchoring → later verification. |
| Dependencies | Identity layer, governance authority records, anchoring model. |
| Security / privacy | Must permit verification without disclosing unrelated source data. |
| Interoperability intent | Stated goal that third parties can verify without ByteShares infrastructure. |
| Maturity | Concept. |
| Source / provenance | Name and framing owner-supplied; specification document not available. |
| Open items | Signature scheme, serialization, canonicalisation, versioning, test vectors. |
A positioning term used in project material for an internet layer built around verified identity and accountable participation. It is framing, not a specification, and nothing on this platform depends on any particular definition of it.
| Element | Current content |
|---|---|
| Purpose | Situate the work: participation that carries verified identity and accountable authority. |
| Scope | Framing only; no normative requirements. |
| Maturity | Positioning term. |
| Source / provenance | Project material; no ratified external definition relied upon. |
| Open items | Whether the term is used publicly at all, and under what definition. |
A conceptual model for accountable AI identity. It treats an AI agent as a governed participant whose identity, curator, permitted memory, behavioural constraints and authority must each be explicit and independently reviewable.
A stable agent identifier links a declared model or service context to its current policy and authority record.
A named curator or institution approves purpose, configuration, source boundaries and lifecycle changes.
Permitted memory sources, retention and disclosure are scoped. Memory does not create authority.
Behavioural constraints describe what an agent may attempt; signed authority determines which consequential actions it may perform. The two are not interchangeable.
| Element | Current content |
|---|---|
| Purpose | Let an institution answer who acted, under whose curation and within which authority, when the actor is an AI agent. |
| Scope | Agent identity, curation, memory boundary, behaviour and authority boundary. |
| Actors | Agent, curator, authorizing body, relying party. |
| Data / proof objects | Agent identity record, curation record, authority grant. Formats to be specified. |
| Lifecycle | Declaration → curation approval → authority grant → operation with evidence → revision or revocation. |
| Dependencies | Identity layer, governance authority, evidence and anchoring. |
| Security / privacy | Memory scope must be bounded, and agent action evidence must not disclose unrelated data. |
| Interoperability intent | Reviewable by institutions outside ByteShares. |
| Maturity | Concept. |
| Source / provenance | Concept owner-supplied; specification document not available. |
| Open items | Identifier format, curation record schema, revocation semantics, enforcement model. |
F.I.P. is conceptual and non-operational. No AI identity registry, agent certification or enforcement service exists.
Each statement on this page is tracked against a source. A claim without a verified source is never presented as fact.
| Claim | Source type | Verification | Note |
|---|---|---|---|
| Sig-56, Web4 and F.I.P. are associated with the ByteShares work | Owner instruction | Verified source on file | The names and framing were provided directly by the project owner. |
| Technical content of Sig-56 and F.I.P. | Specification documents | No source on file | No specification has been supplied for inspection; the structures above are owner-asserted concepts. |
| Any external standards status, adoption or certification | Standards body records | No source on file | None claimed or held. |