AI Governance Institute
← News
Research2026-10-27

Google Cloud's Agent Compute Engine Exposes Non-Human Identity Gap

What happened

Google Cloud published Introducing Agent Compute Engine and expanding agent governance on October 22, 2026. The post describes a new infrastructure product and enterprise governance controls for agentic AI. The guidance calls out agent registries, unique non-human identities for each agent, least-privilege access scoping, centralized policy enforcement, and auditable records for asynchronous agent actions as foundational requirements. Agents interact dynamically with other agents, tools, models, skills, and data. Existing IT inventory and access control programs were not built to capture that. Google frames these controls as necessary for that reason. The publication follows a pattern of major cloud vendors formalizing agentic governance requirements after real-world agent incidents. Those incidents include the 55% of Organizations Run System-Changing AI Agents Without Identity Controls finding and the 1Password Launches Agent Login Controls, Exposing Identity Gap at Scale disclosure.

Why it matters

  • ·Most enterprise AI inventories track human-facing tools, not the non-human identities that agents use to take actions. Without a dedicated agent registry, compliance teams cannot demonstrate to regulators or auditors that access controls cover every active agent. This is now a named expectation in guidance such as the Five Eyes Guidance on the Careful Adoption of Agentic AI Services.
  • ·Agents that interact with other agents can accumulate permissions beyond what any single deployment review approved. Least-privilege access must be enforced at runtime, not just at the time an agent is first deployed, or a single compromised agent can escalate access across the entire environment.
  • ·Asynchronous agent actions, where an agent completes a task long after a human approved the initial instruction, create a gap in audit trails. Logs must capture what each agent did, when, and on whose authority. Without that, organizations cannot meet the audit readiness standard that banking supervisors, the EU AI Act (Regulation (EU) 2024/1689), and state-level laws are beginning to enforce.

Stay ahead of stories like this

Get every Global AI governance development like this one, plus the rest of the week's developments. Every Thursday.

Free every Thursday. Unsubscribe anytime.

Governance controls affected

What to do now

  • ☐Ask your engineering team to provide a complete list of every AI agent currently running in production, including agents that operate on schedules or in the background without direct user interaction, and confirm whether each agent has its own unique identity credential rather than sharing a human user account.
  • ☐Review whether your access control program sets specific, narrow permissions for each agent at the point it is deployed, and verify that those permissions cannot expand automatically when the agent interacts with other agents or tools.
  • ☐Confirm that your audit logging captures agent-to-agent actions, not only actions that a human directly triggers, so that any agent activity can be traced back to an authorization decision in the event of a regulatory inquiry or incident investigation.
  • ☐Update your AI system inventory process to require that every new agent deployment is registered before go-live, with a named business owner, a defined task scope, and a record of what systems and data the agent is permitted to access.
  • ☐Brief your legal and compliance leads on the vendor's agent governance guidance and compare its control expectations against what your current AI governance program actually documents, flagging any gaps for remediation before the next internal audit cycle.

What to watch next

Banking supervisors and data protection authorities are actively asking enterprises to prove that AI agents can be audited, a trend that will accelerate as agentic deployments scale. The EU AI Act (Regulation (EU) 2024/1689) enforcement program is expanding. The CCPA Regulations on Automated Decisionmaking Technology, Risk Assessments, and Cybersecurity Audits already require organizations to account for automated systems that affect consumers. Compliance teams should monitor whether Google Cloud's agent registry and non-human identity expectations become a de facto procurement standard. AWS, Microsoft, and other major cloud providers are publishing similar control frameworks in parallel.

Related Coverage

Research2026-11-12

WEF Report Sets Systemic Risk Baseline for Agentic AI Governance

The World Economic Forum published 'Governing Agentic AI: Managing Systemic Risk for Trusted Adoption' in November 2026. It proposes a governance framework for AI agents operating autonomously across organizations and systems. The report calls for verifiable agent identities, named human or legal owners for each agent, least-privilege access credentials, comprehensive action logging, and incident and near-miss reporting. These recommendations span AI inventories, identity and access controls, third-party risk management, operational resilience, and cross-organizational incident response.

Research2026-11-12

KPMG and Cequence Set Runtime Controls as the Agentic Governance Baseline

Cequence Security and KPMG in Singapore have published a joint report, "Delegating Outcomes, Not Authority: Agentic AI Runtime Security, Governance, and Control." The report argues that static permissions alone are insufficient to govern AI agents. The report recommends runtime controls that evaluate each agent action before it executes, covering least-privilege access, non-human identity, tool-use monitoring, audit trails, and prompt injection defense. Cequence field data from October 2026 tracked more than 281,000 AI agent visits across five industries in a single week.

Corporate Policy2026-11-15

Google Cloud's Policy MCP Server Lets AI Agents Rewrite Cloud Security Rules

Google Cloud has made its Organization Policy Model Context Protocol (MCP) server generally available. The server allows autonomous AI agents and automated pipelines to inspect, evaluate, and create policies and custom constraints across a Google Cloud environment. Enterprises connecting agents to this server must carefully limit what those agents are permitted to change, and log every evaluation and modification request.