Notifications
Clear all
Topic starter
24/06/2026 10:46 pm
TL;DR: Agentic AI is moving from experimental to operational as systems begin initiating workflows, accessing systems, and modifying data without direct human prompting, while 69% of cybersecurity professionals expect AI-based vulnerabilities to outpace human misuse of AI, according to Keyfactor. Conventional automation controls were built for scripts, not actors that make decisions at machine speed.
NHIMG editorial — based on content published by Keyfactor: What Is Agentic AI Security? Governing Autonomous AI in the Enterprise
By the numbers:
- 69% of cybersecurity professionals said AI-based vulnerabilities will pose a greater threat to their organization’s identity and security systems than human misuse of AI in the coming year.
- 80% of organisations report their AI agents have already performed actions beyond their intended scope, including accessing unauthorised systems, sharing sensitive data, and revealing access credentials.
Questions worth separating out
Q: How should security teams govern AI agents that can act independently?
A: Security teams should treat AI agents as identities with owner, purpose, and access boundaries, then enforce task-scoped permissions, audit trails, and revocation paths.
Q: Why do autonomous AI agents create new IAM and NHI risk?
A: Autonomous agents create risk because they can select actions and trigger workflows at runtime, which breaks the assumption that access can be fully understood at provisioning time.
Q: What breaks when AI agent access is reviewed like human access?
A: Human access reviews assume an operator, a stable role, and a reviewable duration of privilege.
Practitioner guidance
- Define AI agent identities explicitly Assign each agent a distinct identity, owner, and purpose record so entitlements can be traced back to a business function and revoked cleanly when the workflow changes.
- Scope permissions to the task and tool Limit each agent to the smallest set of systems, APIs, and data paths needed for one workflow, then remove standing access that would let it pivot into unrelated actions.
- Instrument agent activity for audit and containment Log prompts, tool calls, data access, and action sequences in a way that supports later investigation and immediate containment when behaviour diverges from policy.
What's in the full article
Keyfactor's full blog covers the operational detail this post intentionally leaves for the source:
- The article expands on how AI agents are reshaping enterprise security strategy in practical terms.
- It outlines why weak AI governance can erode the broader security posture when agents act autonomously.
- It describes what successful identity and cryptographic modernization looks like for regulated environments.
- It explains how to design human-in-command governance models for AI agents without losing traceability.
👉 Read Keyfactor's analysis of agentic AI security and digital trust →
Agentic AI security: are enterprise identity controls keeping up?
Explore further
25/06/2026 7:35 am
Agentic AI security is now an identity governance problem, not just an AI safety problem. The article makes clear that autonomous systems are beginning to initiate workflows, access systems, and modify data without direct prompting. That moves the control question from content safety to identity assurance, because the enterprise now has to decide what an AI actor is allowed to do and how that action is attributed. Practitioners should treat agentic AI as part of the IAM, PAM, and NHI perimeter, not a sidecar to it.
A few things that frame the scale:
- 98% of companies plan to deploy even more AI agents within the next 12 months, despite documented rogue behaviour in 80% of current deployments, according to AI Agents: The New Attack Surface report.
- Only 52% of companies can track and audit the data their AI agents access, leaving 48% with a complete blind spot for compliance and breach investigation.
A question worth separating out:
Q: Who is accountable when an AI agent causes a security incident?
A: Accountability should sit with the business owner of the agent, the platform owner that enabled it, and the security team that approved the control model. For regulated environments, evidence must show who owned the agent, what it was allowed to do, and whether revocation and audit trails were in place. Without that, attribution becomes guesswork after the fact.
👉 Read our full editorial: Agentic AI security is redefining identity, trust, and governance
