Accountability layer for AI agents
Identity is not accountability.
Every existing standard answers whether an agent is cryptographically who it claims to be. None answers which legal entity pays when it causes harm.
IDN is a cross-jurisdictional layer, backed by a federation of regulated issuers, whose accountability assertion is KYC-grade, portable across organizational boundaries, verifiable by a counterparty with no relationship to the sponsor, and legally meaningful rather than merely access control.
See who is behind an agent
A verifier asks one question and gets a signed answer. Pick a sponsor handle to query the attestation API.
No attestation found.
An agent crosses an organizational boundary
An autonomous agent acts on one company's behalf against a system it does not own: a payment, a contract change, a booking, a filing.
The receiving party cannot establish who is accountable
The counterparty can confirm the agent holds a key. It cannot establish which KYC-verified legal entity stands behind it, or who bears liability when the action causes harm.
The workaround is a human, not a design
Today the gap is filled by a person pausing to confirm. That is not a security design. It is the absence of one, and it does not scale to agents acting continuously across organizations.
One question, one signed answer, across the boundary.
- 01
Issuer verifies the sponsor
An accredited issuer runs KYC on the sponsor and records the legal entity, its jurisdiction, and its registry identifier.
- 02
Sponsor binds the agent
The sponsor binds a specific agent to itself under a scoped, short-lived attestation the issuer signs.
- 03
Agent presents its identifier
When the agent acts across a boundary, it presents its identifier to the counterparty.
- 04
Verifier resolves and receives a signed answer
The counterparty resolves the identifier and gets a signed statement naming the accountable entity, verifiable offline with no relationship to the sponsor.
The obligations are dated, and one is already in force.
- 2 August 2026
EU AI Act, Article 50
Transparency obligations for AI systems apply. Penalties reach EUR 15 million or 3 percent of turnover.
- 18 January 2027
GENIUS Act
Takes effect by this date at the latest.
- 2 December 2027
EU AI Act, high-risk
High-risk obligations apply, following the enacted Digital Omnibus.
Working with the frameworks that regulate agents
IDN extends what is already converging in agent identity. Every handle is a W3C DID at the base. Every attestation is a JWS. Bridges to GLEIF vLEI, ISO 15118-20, and OpenID AIIM are first-class concerns.
- W3C DID Core
- W3C VCDM 2.0
- JWS · RFC 7515
- Ed25519 · RFC 8037
- ML-DSA-65 · FIPS 204
- OpenID AIIM
- GLEIF vLEI
- ISO 15118-20