Meet AI technical requirements

Understanding the technical requirements that apply across the AI lifecycle and how to translate them into practical design, delivery and operational decisions.

Meeting technical requirements is not a one-off compliance activity. Agencies should consider architecture, data, security, privacy, testing, human oversight, accessibility, deployment and monitoring from the earliest stages of an AI use case and throughout its operation.

Start with the AI technical standard

The AI technical standard provides a consistent approach to the design, development, deployment and operation of AI systems across government. Use the standard to identify the technical considerations, evidence and controls that may be needed for your use case. Apply it alongside the AI policy, your agency's governance arrangements and relevant whole-of-government requirements.
 

Using agentic AI?

Agentic AI systems may act with greater autonomy and can introduce additional technical, governance and operational risks. Use the Agentic AI addendum alongside the AI technical standard to understand the additional considerations that apply.

Apply requirements across the AI lifecycle

Plan and assess
Define the intended purpose, users and operating context. Assess whether AI is appropriate and identify potential impacts, risks, constraints and applicable requirements before progressing.
Design and develop
Translate requirements into architecture, data, security, privacy, accessibility, human oversight and system design decisions. Document assumptions, dependencies and controls.
Test and evaluate
Establish clear measures and test the system's performance, reliability, safety and behaviour before deployment. Include testing for foreseeable failure modes and unintended impacts.
Deploy and integrate
Confirm the system is ready for its intended environment, users and level of risk. Plan for secure integration, access controls, change management, incident response and operational support.
Monitor and improve
Monitor performance, risks and impacts over time. Reassess the system when its use, data, model, environment or risk profile changes, and maintain appropriate records throughout its lifecycle.
Decommission and retire
Safely retire, shut down or repurpose an AI system using a structured approach while managing security, data and compliance risks.

Supporting resources

Where do I go next?

Technical requirements should be considered alongside policy, governance, capability and ongoing assurance.

Assess AI use cases
Identify impacts, risks, value and suitability before committing to an AI solution.
Meet accountability and transparency requirements
Make sure governance, accountability, documentation and transparency arrangements support the use case throughout its lifecycle.
Implement AI safely
Use assessment tools to identify risks, understand impacts and support the responsible use of AI.
Build capability
Develop the technical, operational and governance skills needed to design, manage and oversee AI effectively.

Connect with the digital community

Share, build or learn digital experience and skills with training and events, and collaborate with peers across government.

Join the Digital Profession