notes: events/talks/kericonf26-d2-01-smith-the-future-of-keri

This data as json

path title type status classification words body
events/talks/kericonf26-d2-01-smith-the-future-of-keri Samuel M. Smith: The Future of KERI (KERIConf 2026) summary v0.1 — 2026-10-15: drafted by Claude from the curated transcript; unread by Mark; marked partner on Mark's word 2026-10-21, paraphrase only, no quotation partner 827 # Samuel M. Smith: The Future of KERI (KERIConf 2026) **Abstract.** In the Day 2 opening plenary, Samuel M. Smith described where he sees KERI (Key Event Receipt Infrastructure) and ACDC (Authentic Chained Data Containers) going next: bridging trust domains, using ACDCs as general authorizations, handling revoked delegation chains through reversion and transfer of registry control, applying the same mechanisms to AI agents, and building accountable digital refuges with TSP (Trust Spanning Protocol). Parts of the talk are proposals for the next ACDC specification version. It is for implementers and specification readers. ## Summary Smith framed his work as an effort to fix the internet, which he came to see required solving identity. He reviewed ACDCs as graph fragments that each authenticate separately, giving perpetually verifiable issuances and delegable, attenuable, aggregable, revocable and transferable authority. With these, ACDCs can carry trust between trust domains without loss, a property he described using Daniel Hardman's phrase lossless trust transfer. Systems that cannot bridge trust domains instead enlarge a single domain, which in his account produces recursive privilege escalation attacks. He then introduced EVAC (Edge Verifiable Access Control): a resource controller grants access to another edge through an ACDC, and mutual authentication at every edge prevents impersonation, unlike a VPN (virtual private network) relying on a shared secret. His proposal, not yet published as a white paper, adds a resource map to the ACDC attribute section, listing routes and capabilities in the manner of object capabilities. He pointed to an issuee registry field already in the specification: if the holder anchors each presentation in a registry it controls, a verifier checks both the issuer and the presenter registries, and the holder can detect theft of its signing keys. The next problem was delegation chains in which a delegate's authority is revoked after it has issued credentials, such as purchase orders. Binding the delegate's key state at grant and at revocation through a bound blinded state registry preserves issuances made while authority held, but only if they expire on their own. For dynamically revocable issuances he proposed reversion of authority: control over those issuances passes back up the edge to the delegator, repeatedly if needed. He further proposed a transfer event for the TEL (Transaction Event Log), analogous to rotation in a KEL (Key Event Log), so control of a registry can move to a new issuer without reissuing credentials, aimed at ACDC version 1.1. Examples included guardianship after divorce. Karla McKenna noted from the audience that this could let GLEIF (the Global Legal Entity Identifier Foundation) take temporary control when a QVI (Qualified vLEI Issuer) leaves, replacing today's fixed grace period. He said asset transfer is a separate matter, to be covered by Ned Smith. He extended EVAC to Edge Verifiable Agentic Control: an AI (artificial intelligence) agent is a delegate like any other, given a delegated AID (autonomic identifier), with no change to the mechanisms. In questions he said limits on delegation depth come from edge operators or an ecosystem governance framework, and that the verifier enforces least privilege. Phil Windley added a distinction between delegation as data and delegation as policy. In the last part he argued that safety and privacy both require accountability, since second parties can always correlate first parties. Privacy comes from covert rather than clandestine means, from end-to-end verifiability, and from only-at-end viewability using hybrid public key encryption. TSP combined with the KERI suite, he said, supports such relationships, the subject of his breakout talk. ## Key terms perpetually verifiable issuance, cross trust domain transfer, lossless trust transfer, recursive privilege escalation attack, EVAC, issuee or usage registry, reversion of authority, transfer event, digital refuge, only-at-end viewability ## Claims and their evidence - The ACDC specification already defines an issuee registry identifier that lets a presenter anchor presentations and detect key compromise. [presenter-reported] - A transfer event in the TEL, carrying only the new issuer AID, would allow control of a registry to change without reissuing its credentials; it is proposed for ACDC version 1.1. [presenter-reported] - Under reversion of authority, revoking a delegate's authority returns control of its dynamically revocable issuances to the delegator. [informative] - Delegation to AI agents uses the same ACDC mechanisms as delegation to people or organizations. [informative] - In the vLEI (verifiable Legal Entity Identifier) system today, credentials from a departing QVI remain valid for a fixed grace period. [presenter-reported] ## Related ACDC specification (version 1.1 proposals), KERI specification, Trust Spanning Protocol, vLEI ecosystem governance framework, KERIpy, KERIConf 2026 talks by Philip Feairheller and Phil Windley on authorization, Ned Smith's Secure Asset Transfer Protocol talk, Samuel M. Smith's breakout on SPAC (Secure Privacy, Authenticity and Confidentiality) and TSP ## Provenance Drafted by Claude on 2026-10-15 from the curated transcript (d2-01-samuel-smith-the-future-of-keri.md), all minutes read (the transcript runs to about 1:03:20, though its header gives the duration as 52:10; a short passage around 47:10 to 48:30 is missing during a microphone change). No quotations; paraphrase only. Not yet read by Mark.

Links from other tables