2026-09-21 AI Agent Identity News
Summary: Dated digest of raw/2026-09-21-ai-agent-identity-news.md. No new Estonia or other-country developments; a Monday research roundup surfaces an authorization-architecture review and two IETF drafts on agent identity and signed authority receipts.
Sources: 2026-09-21-ai-agent-identity-news.md
Last updated: 2026-09-21
Section 1 — Estonia
No new non-duplicate Estonia sources were found. The report’s own search notes confirm it surfaced only the June proposal, RIA’s Aruait project page, and the 8 September correction — all already recorded on estonia-ai-agent-id. No change made.
Section 2 — Other countries
No new independent country-level policy, government, regulatory, or major-news source was found. The report explicitly notes the ITU–World Bank TIDA focus group (covered on 2026-09-18-ai-agent-identity-news) was not repeated. No change made to any country page.
Section 3 — Research (Monday roundup)
Three items, all newly surfaced this run:
- Authorization Architectures for Tool-Using AI Agents — an academic review of 89 primary sources organizing agent authorization around a principal chain (human user → operator/deployer → orchestrator agent → sub-agent → tool endpoint) and five control layers, flagging the authorization decision at tool-invocation time as a major unresolved point (source: 2026-09-21-ai-agent-identity-news.md, citing arXiv:2609.15906, published 2026-09-14).
- IETF WIMSE working-group draft, “AI Identity Management System” (draft-ietf-wimse-aims-00) — composes existing mechanisms (unique workload identifiers, short-lived credentials, OAuth-based delegated authorization, transaction tokens, cross-domain access, monitoring, policy, compliance) into an agent identity-management stack; now a working-group document rather than an individual submission (source: 2026-09-21-ai-agent-identity-news.md, citing IETF Datatracker, published 2026-09-15).
- “Agent Authority Transition Receipts for Agentic Systems” Internet-Draft — proposes a signed, non-bearer receipt binding one proposed operation to the represented principal, acting agent, action digest, policy/version, evaluated evidence, validity period, decision, and predecessor authority state, with fail-closed handling of missing or unverifiable evidence (source: 2026-09-21-ai-agent-identity-news.md, citing IETF Datatracker, published 2026-09-12, updated 2026-09-14).
Full analysis and cross-links added to research, agent-standards-and-interoperability, and agent-authorization-and-delegation.