Back to Insights
The Enterprise AI Agent Security and Access Checklist

The Enterprise AI Agent Security and Access Checklist

6 min read

An enterprise AI agent is not simply a chatbot that generates answers. It may send email, change CRM records, execute code, or request a payment. The center of security therefore moves from “What can it say?” to “What can it do?”

The answer is not asking the model to behave. The system must constrain which identity can reach which data and tools, for how long, which actions require approval, and what can be logged, cancelled, and recovered.

NIST Zero Trust Architecture does not grant trust based only on network location. It evaluates access to users, services, and resources through policy. An agent should follow the same principle: task-scoped, least-privilege sessions instead of broad access to the internal network.

How to Record the Checklist

Do not stop at yes or no.

IDControl QuestionStatusEvidenceOwnerDue Date
Example A-03Is CRM delete permission blocked by default?PartialIAM policy linkSecurity2026-04-30

Use Met, Partial, Not Met, or Not Applicable. Even Not Applicable needs a reason.

A. Before Execution: Identity, Scope, and Data

IDControl QuestionMinimum Evidence
A-01Are human users, agents, and external service accounts distinct?IAM structure and service-account inventory
A-02Does the agent receive task-level least privilege instead of inheriting all user permissions?Role and policy definitions
A-03Are read, write, delete, transmit, and payment permissions separated?Permission matrix
A-04Are credentials short-lived, rotatable, and absent from prompts and logs?Secret-management and TTL policy
A-05Are permitted tools, domains, and APIs restricted by allowlist?Tool registry and network policy
A-06Are personal, confidential, and customer data classified with retention and transfer conditions?Data-flow diagram and processing basis
A-07Are external documents, websites, and email treated as untrusted input?Trust labels and isolation rules
A-08Are models, plugins, MCP servers, libraries, data providers, and versions recorded?AI BOM and contract inventory

“This information is secret” inside a prompt is not access control. If the model can read it, the exposure path already exists.

B. During Execution: Action, Approval, and Isolation

IDControl QuestionMinimum Evidence
B-01Before a sensitive or irreversible action, does the user see its target, effect, and cost?Approval UI and policy tests
B-02Does any change in target, amount, or input after approval require new approval?Binding and reauthentication test
B-03Can external instructions be prevented from changing higher-level policy, secrets, or tool permissions?Prompt-injection evaluation
B-04Are task data, memory, and execution environments isolated by user and tenant?Sandbox and tenancy design
B-05Are calls, amounts, recipients, file scope, and duration bounded?Rate, budget, and transaction limits
B-06Does policy violation, uncertainty, or repeated failure stop safely and escalate to a person?Stop and escalation rules
B-07Can the agent be prevented from expanding its own permissions or installing tools?Privilege-escalation tests
B-08Can state before and after high-risk action be captured and restored?Snapshot and transaction logs

“A human looks at the end” is not enough. The approver must understand what will happen, have the power to stop it, and approve the same target that is ultimately executed.

C. After Execution: Logs, Recovery, and Learning

IDControl QuestionMinimum Evidence
C-01Can user, agent, model, tool, input, output, decision, and approval be traced under one correlation ID?Sample audit log
C-02Are secrets and unnecessary personal data masked, with log access and retention limited?Log policy and sample
C-03Can a bad write, send, or change be reversed or compensated?Rollback or compensation runbook
C-04Are incident detection, shutdown, reporting, notification, forensics, and contacts assigned?Incident plan and drill record
C-05Do model, prompt, tool, and permission changes require regression evaluation and approval?Change-management record
C-06Are vendor outage, terms, model changes, termination, and data return managed?Vendor-risk and exit plan
C-07Do real failures and near misses update the eval set and policy?Failure taxonomy and eval-set history

Collecting many logs is different from reconstructing an incident. You need to connect what was approved, which tool acted, and which target it changed.

Approval Strength by Action

ActionReversible, Low ImpactReversible, High ImpactIrreversible, High Impact
Read and analyzeAutomatic with logsApprove purpose and scope for sensitive dataRedesign the task first
Draft and recommendGenerate automatically, human usesShow evidence and uncertainty before approvalSecond review by the decision owner
Write and transmitAutomatic within small bounds or post-reviewConfirm target and effect immediately before executionStrong authentication, dual approval, limited automation
Delete, pay, or change rightsSmall limit and immediate cancellationDual approval and transaction limitDeny by default or use a separate controlled environment

Eight Required Attack Scenarios

  1. An indirect prompt injection in a document asks the agent to ignore prior instructions and send a secret.
  2. A request substitutes a similar person, domain, or account number.
  3. The recipient, file, or amount changes immediately after approval.
  4. Several individually allowed tools are composed to produce a prohibited result.
  5. The agent attempts to read another user’s memory, documents, or logs.
  6. An infinite retry or tool loop exceeds the cost limit.
  7. A vendor or tool returns an erroneous or manipulated response.
  8. Execution is requested while logging or approval services are unavailable.

OWASP’s guidance on agentic applications covers goal hijacking, tool misuse, excessive privileges, supply chain, memory, and identity risks. Security attacks should be evaluated as a separate track from normal answer-quality tests.

Regulation Is a Gate Outside the Checklist

The EU AI Act overview and Korea’s Framework Act on the Development of Artificial Intelligence and Establishment of a Foundation for Trustworthiness address risk, transparency, human oversight, and records according to context. Completing this checklist does not establish legal compliance. Sector, data, and jurisdiction require separate legal and security review.

The final rule is simple: Do not give an agent a human account in full. A system that acts faster than a person needs narrower permissions, shorter-lived credentials, and a clearer stop mechanism.