NSW has released Australia’s first guidelines for agentic AI to help government agencies use the technology safely and ethically. The guidelines aim to support AI systems that can make decisions and plan actions, rather than just follow rules.
The guide provides checklists and advice for agencies to determine when an AI agent is needed. It outlines how to maintain human oversight, conduct regular reviews of AI outputs, and apply privacy and security controls.
The guidance was developed with input from AI experts and chief architects. It helps agencies set policies and assurance processes before piloting or deploying AI systems. High-risk AI projects must be reviewed by the AI Review Committee.
Agentic AI is already in early use in the private sector. Some banks use AI to detect fraud, while online stores use AI agents to help with product descriptions and digital layouts. NSW agencies will evaluate AI projects using the updated AI Assessment Framework.
The Department of Customer Service will continue refining the guidance as global standards for AI evolve. The AI agent guide is available online for public access.