2026-10-19 AI Agent Identity News
Summary: Dated digest of raw/2026-10-19-ai-agent-identity-news.md (Monday). No new Estonia or other-country news; three research items on action-bound approval, evidence-graded agent maturity, and continuing governance for general agents.
Sources: 2026-10-19-ai-agent-identity-news.md
Last updated: 2026-10-19
Section 1 — Estonia
No new relevant, non-duplicate source. RIA’s 15 October Aruait update and ERR’s 15 October legal commentary were already covered (2026-10-18-ai-agent-identity-news); crawl dates were not treated as publication dates (source: 2026-10-19-ai-agent-identity-news.md).
Section 2 — Other countries
No new development after the latest saved report’s cutoff. CERT-In, HKMA, FSB, ICO and NIST items excluded as duplicates; reporting on disclosures dated 16 October predates the window (source: 2026-10-19-ai-agent-identity-news.md).
Section 3 — Research
Three primary preprint/journal items (peer review not established for the preprints); details on research:
- Approve What, Exactly? — Kodathala, Pollishetty, Cargill (arXiv 2610.21063, 15 Oct). Audit of ten runtimes found no third-party-verifiable approval record and post-approval argument changes in four of eight where measurable; Agent Step-Up gateway binds a passkey signature to a canonical action digest (source: 2026-10-19-ai-agent-identity-news.md).
- AgentStripes — Naik et al. (arXiv 2610.18603, 13 Oct). Evidence-graded maturity model, 40 capabilities in 12 areas; exploratory static application to 30 public repos; authors say it does not validate the model (source: 2026-10-19-ai-agent-identity-news.md).
- Risk for AI Agents — Faroldi, European Journal of Risk Regulation (28 Sept; DOI 10.1017/err.2026.10152). Conceptual argument that general agents need governance beyond product-style risk assessment (source: 2026-10-19-ai-agent-identity-news.md).
Contradictions
None.