HKMA and Cloud Vendors Turn Agent Identity Into Continuous Access Control

Regulators and cloud vendors are converging on a model where AI agents need their own identities, continuous authorization, and auditable accountability.

Updated

What is this trend?

Banks and cloud vendors are turning AI agent identity into continuous access control, so every delegated action can be verified, scoped, monitored, and revoked.

  • Distinct agent IDs are replacing shared service accounts and opaque runtime access.
  • Authorization is shifting from one-time setup to continuous checks on scope, duration, and drift.
  • Audit trails now need to tie each agent action to a human or institution sponsor.
  • Cloud platforms are adding identity propagation, short-lived credentials, and per-request enforcement.
  • Security teams must connect IAM, SIEM, and AI governance into one control loop.

What’s the latest?

How it developed

  1. Inference Cost Optimization, Runtime Agent Security, and Cloud Sovereignty Constraints
  2. Overprivileged AI runtimes expose security risk, governed remediation shifts IT from triage to policy
  3. AI Capacity Becomes a Portfolio, Sovereign AI Goes Deployable, and Observability Adds Governance

Go deeper

Curated long-form picks on this trend — podcasts, videos, and analysis, by seniority.

Related trends

Stay ahead in Information Technology (IT)

Get the weekly Information Technology (IT) brief in your inbox — the developments, what they mean by seniority, and what to do next.