AI Governance Institute
← News
Research2026-10-09

55% of Organizations Run System-Changing AI Agents Without Identity Controls

Source

JumpCloud

What happened

JumpCloud published survey findings showing that more than half of organizations deploy or pilot AI agents that can alter systems, change permissions, update records, or trigger workflows. Despite that reach, 59 percent of respondents had not fully applied their human access management policies to those agents. The report focuses on five control gaps: non-human identity inventories, accountable ownership, limited credentials, authorization reviews, and access revocation. The findings echo a pattern documented across recent incidents. Both the AI Agents Running as Users: Rig Security's $12M launch story and the Identity Controls Are Necessary for AI Agents, But Not Sufficient analysis highlighted the same structural gap. Organizations manage human access differently than they govern agent access. Regulators and auditors are increasingly treating agent identity as a foundational control, not an optional one.

Why it matters

  • ·Organizations subject to the EU AI Act (Regulation (EU) 2024/1689) or financial-sector model risk rules face direct exposure. Agents that can change permissions or records without a documented access trail are likely to fail conformity assessments and supervisory reviews. The 59 percent figure gives regulators a concrete benchmark for how widespread the gap is.
  • ·The operational risk is immediate: an agent running with broad, unreviewed permissions can modify data, escalate its own access, or take irreversible actions before any human is aware. Without a revocation procedure, even a decommissioned agent may retain live credentials and continue acting.
  • ·Accountability is the hardest gap to close. If no named owner is recorded for an agent's identity and permissions, compliance teams cannot answer basic audit questions. They cannot establish who approved the access, who can revoke it, or who is responsible when something goes wrong.

Stay ahead of stories like this

Get developments like this, plus everything else that matters in AI governance. Every Thursday.

Free every Thursday. Unsubscribe anytime.

Governance controls affected

AGT-009Agent and Non-Human Identity ManagementAGT-001Agent Permission BoundariesAGT-027Agent Ownership and Accountability RegisterAGT-025Agent Access Entitlement Runtime Re-ValidationAGT-005Human Approval Gate for Irreversible Agent Actions

What to do now

  • ☐Pull a complete list of every AI agent in use or in testing across the organization, noting whether each one has the ability to change systems, move money, update records, or modify access for other users or systems.
  • ☐For each agent identified, confirm that a named human owner is recorded in your access management system and that the agent's credentials are separate from any individual employee's login.
  • ☐Audit whether your existing access management policies, including periodic reviews and revocation procedures, have been formally extended to cover AI agents, and document any gaps in writing.
  • ☐Require that any agent with the ability to change permissions or take irreversible actions goes through a human approval step before those actions execute, and verify that this gate is technically enforced, not just a policy statement.
  • ☐Set a recurring review cycle for agent credentials, no less frequent than your human access reviews, and confirm that credentials are automatically revoked when an agent is decommissioned or its scope changes.

What to watch next

Regulatory guidance on non-human identity controls for AI agents is accelerating. The Five Eyes Guidance on the Careful Adoption of Agentic AI Services already names identity and least-privilege as baseline requirements. Financial regulators in the EU and US are incorporating agent access controls into model risk supervisory reviews. Enforcement actions targeting agents that operated with unchecked permissions are expected to increase through 2027. Compliance teams should also monitor updates to EU AI Act (Regulation (EU) 2024/1689) conformity guidance. If non-human identity is addressed explicitly, the current 59 percent gap becomes a formal compliance deficiency, not a best-practice shortfall.

Related Coverage

Corporate Policy2026-11-02

Azure Telemetry Confirms Prompt Injection as the Top Agent Attack. Accountability Is the Fix.

Jones Walker published guidance drawing on Azure AI workload telemetry. It found that malicious instructions injected through model and tool outputs are the most common observed attack against enterprise AI agents. The article recommends that every deployed agent carry a named human sponsor, auditable identity metadata, and time-limited access rights. It also calls for recertification whenever a material aspect of the agent changes.

Research2026-10-27

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

Google Cloud published implementation guidance for enterprise agentic AI governance, introducing Agent Compute Engine alongside controls for agent registries, unique agent identities, least-privilege access, and centralized policy enforcement. The guidance addresses how dynamic interactions among agents, tools, models, and data create compliance blind spots. Organizations running AI agents in production need to assess whether their current AI inventories and access controls cover non-human identities and agent-to-agent activity.

Corporate Policy2026-10-29

Microsoft CISO Guidance Makes Agent Identity and Observability the Enterprise Baseline

Microsoft has published guidance for enterprise security leaders on governing AI agents in production environments. It calls for assigning agents their own managed identities and limiting their access to only what each task requires. Security teams must maintain continuous visibility over what each agent owns, can reach, and has done. It directly affects identity management, access controls, audit logging, and incident response programs.