Government Advisory
Get the briefing
Government Advisory

ASD Publishes Guidance on How Agentic AI Harnesses Shape Security and Risk

ACSC · 22 Sept 2026
Key Takeaway If your business is adopting agentic AI tools, review the ASD guidance first and ask your vendors how their harness handles security and governance.

The Australian Signals Directorate (ASD), through the Australian Cyber Security Centre (ACSC), has released new guidance on agentic AI harnesses. The guidance aims to help organisations understand how the harness influences security, governance and operational risk in agentic AI systems.

Agentic AI refers to systems that can act with some independence to complete tasks. The harness is the surrounding framework that supports the AI model in operating this way. According to the ACSC summary, the design of that harness matters for how safe and well controlled these systems are, not just the AI model itself.

The summary available to us does not set out the specific recommendations in the guidance. Businesses that are using or considering agentic AI tools should read the full publication on the ACSC website to see what ASD advises on governance and risk management.

ASD ACSC agentic AI AI governance risk management
Read original source -> Get the briefing Get the news API, A$10 a month
Running AI agents? Eight controls for securing them, mapped to the ISM ->

Summarised by CISO AI from ACSC, written with Claude Sonnet 5.5. We link back to every original so you can read it yourself.