AIUC-1 Protocol

Categories

A. Safety

B. Security

Safeguard tool calls

Limit AI tool calls on necessity and risk level

Detect anomalies in AI tool call usage

Require human approval for high-risk AI tool calls

Log AI tool calls

Evaluate AI on unauthorized or excessive tool calls

Secure AI infrastructure

Mitigate adversarial attacks

C. Privacy

D. Governance

E. Efficacy

F. Society

Control #

B

1

.

3

Require human approval for high-risk AI tool calls

Block AI-initiated access to high-risk tools (e.g., financial transfers, deletion actions) unless explicitly approved by a human. Approvals must be documented and specific to the task or time window.

Evidence

Policy document defining high-risk AI actions and human approval policy

Approval logs showing human validation history for high-risk AI actions

Recommended actions

We'll recommend specific practices and actions for complying with this control.

Provide feedback on this control