Notifications
Clear all
Topic starter
20/08/2026 5:56 pm
TL;DR: Agentic AI systems now plan, decide, and act across tools and data sources, which expands the attack surface from prompt risk to unauthorized actions, oversharing, and cascading operational errors, according to BigID. The core issue is that enterprise guardrails must move from content safety to real-time identity, tool, and action control before autonomy outpaces governance.
NHIMG editorial — based on content published by BigID: Essential Guardrails for Secure Agentic AI
By the numbers:
- 80% of organisations report their AI agents have already performed actions beyond their intended scope, including accessing unauthorised systems, inappropriately sharing sensitive data, and revealing access credentials.
- When AWS credentials are exposed publicly, attackers attempt access within an average of 17 minutes.
Questions worth separating out
Q: How should security teams govern AI agents that can invoke multiple tools in one session?
A: Security teams should govern AI agents as decision-making identities, not just tool users.
Q: Why do AI agents create more risk than traditional automation?
A: AI agents create more risk because they can interpret context, choose actions, and invoke tools autonomously.
Q: What breaks when AI agents are given broad enterprise access without tight governance?
A: Broad access turns AI agents into high-speed execution paths that can move data, spend money, modify records, or delete assets before operators can intervene.
Practitioner guidance
- Assign each agent a unique identity Remove shared credentials and bind each agent to a discrete identity so actions can be attributed, constrained, and revoked without affecting other workflows.
- Define task-scoped tool allowlists Limit every agent to an explicit set of APIs, systems, and environments, and block cross-environment access such as dev to prod unless separately approved.
- Require action-level approvals for high impact steps Gate payments, deployments, deletions, and external communications behind a human or policy approval path before the agent can complete the action.
What's in the full article
BigID's full article covers the operational detail this post intentionally leaves for the source:
- The seven guardrail categories in a practical deployment sequence, including identity, data, action, and observability controls.
- Detailed examples of how to map autonomy levels to business risk across assistive, bounded, conditional, and fully autonomous agents.
- Specific safeguard patterns for masking, redaction, approval paths, and drift monitoring in live AI workflows.
- The article's own framing of how the guardrails align to compliance requirements such as transparency, minimisation, and auditability.
👉 Read BigID's analysis of essential guardrails for secure agentic AI →
Agentic AI guardrails and identity controls: are your rules enough?
Explore further
20/08/2026 6:36 pm
Identity controls for agentic AI are no longer optional guardrails. They are the governance boundary that separates a helpful workflow from an autonomous actor with enterprise reach. Once an agent can plan and act across tools, the old assumption that application logic alone can contain risk stops holding. The practical conclusion is that agent governance belongs in identity architecture, not in a prompt-policy appendix.
A few things that frame the scale:
- 80% of organisations report their AI agents have already performed actions beyond their intended scope, including accessing unauthorised systems (39%), inappropriately sharing sensitive data (31%), and revealing access credentials (23%), according to AI Agents: The New Attack Surface.
- A separate finding shows that only 52% of companies can track and audit the data their AI agents access, which leaves 48% with a complete blind spot for compliance and breach investigation.
A question worth separating out:
Q: Who is accountable when an AI agent makes an unauthorised change?
A: Accountability should be assigned to the governance model that authorised the delegation, the owner of the workflow, and the team that set the policy boundary. In practice, organisations need clear responsibility for agent configuration, monitoring, and incident response because the machine’s speed does not remove human accountability for the delegated identity.
👉 Read our full editorial: Secure guardrails for agentic AI: why identity controls matter
