Direct link: www.digital.gov.au/ai-tech-standard
Responsible agency: Digital Transformation Agency
Last updated: 30 July 2025
The Technical standard for government’s use of artificial intelligence brings together a set of practices for designing, developing, deploying and using AI systems. It reinforces the Australian Government’s AI ethics principles by embedding fairness, transparency and accountability into a set of technical requirements and guidelines.
This standard complements the Policy for responsible use of AI in government, the Pilot Australian Government artificial intelligence assurance framework, and the Voluntary AI safety standard.
Access the standard
The digital.gov.au website hosts the Technical standard for government’s use of artificial intelligence (full text).
Apply the standard
The standard uses a reference AI lifecycle model to ensure holistic coverage of an AI system from inception to retirement.
The following sections provide an overview of the requirements for agencies delivering and operating AI solutions, arranged as a set of statements and organised by the stages of this lifecycle model.
Criteria for each statement, and guidance on how agencies can comply with these requirements, are included in the full text of the Technical standard for government’s use of artificial intelligence.
Whole of AI lifecycle
Whole of AI lifecycle requirements incorporate challenges and considerations that apply across multiple AI lifecycle stages.
- Statement 1: Define an operational model
- Statement 2: Define the reference architecture
- Statement 3: Identify and build people capabilities
- Statement 4: Enable AI auditing
- Statement 5: Provide explainability based on the use case
- Statement 6: Manage system bias
- Statement 7: Apply version control practices
- Statement 8: Apply watermarking techniques
Stage 1: Design
The design stage includes concept development, requirements engineering and solution design.
- Statement 9: Conduct pre-work
- Statement 10: Adopt a human-centred approach
- Statement 11: Design safety systematically
- Statement 12: Define success criteria
Stage 2: Data
The data stage involves establishing the processes and responsibilities for managing data across the AI lifecycle. This includes data used in experimenting, training, testing and operating AI systems.
- Statement 13: Establish data supply chain management processes
- Statement 14: Implement data orchestration processes
- Statement 15: Implement data transformation and feature engineering practices
- Statement 16: Ensure data quality is acceptable
- Statement 17: Validate and select data
- Statement 18: Enable data fusion, integration and sharing
- Statement 19: Establish the model and context dataset
Stage 3: Train
The train stage covers the creation and selection of models and algorithms. The key activities in this stage include modelling, pre- and post-processing, model refinements, and fine-tuning. It also considers the use of pre-trained models and associated fine-tuning for the operational context.
- Statement 20: Plan the model architecture
- Statement 21: Establish training environment
- Statement 22: Implement model creation, tuning and grounding
- Statement 23: Validate, assess and update model
- Statement 24: Select trained models
- Statement 25: Implement continuous improvement frameworks
Stage 4: Evaluate
The evaluate stage involves the testing, verification and validation of the whole AI system. It is assumed that agencies have existing capability on test management and on testing traditional software and systems.
- Statement 26: Adapt strategies and practices for AI systems
- Statement 27: Test for specified behaviour
- Statement 28: Test for safety, robustness and reliability
- Statement 29: Test for conformance and compliance
- Statement 30: Test for intended and unintended consequences
Stage 5: Integrate
The integrate stage of the AI lifecycle focuses on implementing and testing an AI system within an agency’s internal organisational environment, including with its systems and data.
- Statement 31: Undertake integration planning
- Statement 32: Manage integration as a continuous practice
Stage 6: Deploy
The deploy stage involves introducing all the AI technical components, datasets and related code into a production environment where it can start processing live data.
- Statement 33: Create business continuity plans
- Statement 34: Configure a staging environment
- Statement 35: Deploy to a production environment
- Statement 36: Implement rollout and safe rollback mechanisms
Stage 7: Monitor
The monitor stage of the AI lifecycle includes operating and maintaining the AI system. Monitoring is critical to ensuring the reliability, availability, performance, security, safety and compliance of an AI system after it is deployed.
- Statement 37: Establish monitoring framework
- Statement 38: Undertake ongoing testing and monitoring
- Statement 39: Establish incident resolution processes
Stage 8: Decommission
The decommission stage of the AI lifecycle focuses on the planning, delivery and documentation of decommissioning activities.
- Statement 40: Create a decommissioning plan
- Statement 41: Shut down the AI system
- Statement 42: Finalise documentation and reporting