Human Oversight

Approval Bypass in AI Agents

The real control is whether approval state is enforced before action, not whether the prompt asks the model to be careful.

An action route bypasses a separate approval checkpoint and reaches a business operation. — Approval must precede action.

Definition

What is Approval Bypass?

Approval bypass occurs when an agent completes, influences, or routes around a required human or system approval.

Agents can draft, summarize, select approvers, prepare payloads, and execute follow-up actions. Weak approval design turns suggestion into action.

Attack Scenarios

Where this shows up in real agent workflows.

The exact exposure depends on authority, connected systems, identity, approval state, and evidence quality.

Scenario

A casual confirmation is treated as authorization.

Scenario

A summary hides material risk from a reviewer.

Scenario

A tool executes before approval status is verified.

Business Impact

Why it matters

Unauthorized actions appear approved.
Accountability blurs across requester, approver, owner, and agent.
Audit evidence does not match the real decision.

Exploit Path

How attackers exploit it

Exploit ambiguous language or stale approval state.
Manipulate evidence shown to the reviewer.
Use a tool path that does not check approval objects.

Detection and Testing

How to detect and test for approval bypass.

Detection signals

Actions occur before approval timestamps.
Approval summaries omit sensitivity, scope, or irreversible effect.
Requester, approver, and executor share the same identity.

Test methods

Test missing, expired, partial, conflicting, and revoked approvals.
Verify tool-side approval enforcement.
Review whether approval context is complete enough for human judgment.

Review adjacent failure paths.

Back to Autonomous AI Threat Library

Unsafe Tool Use

Unsafe tool use occurs when an agent invokes a legitimate tool with the wrong purpose, parameters, timing, destination, or authority.

Explore topic

Excessive Permissions

Excessive permissions occur when an agent can read, decide, approve, or change more than the business process requires.

Explore topic

Audit Gaps

Audit gaps occur when teams cannot reconstruct why an agent acted, which control applied, who owned the decision, or what evidence supports the outcome.

Explore topic