Western Australia AI Policy Framework

Overview

The Western Australia AI Policy Framework, enacted on March 5, 2024, serves as a foundational guide for the ethical, safe, and responsible development and deployment of Artificial Intelligence (AI) within the Western Australian public sector. This comprehensive framework is spearheaded by the Office of Digital Government, a division of the Department of the Premier and Cabinet, underscoring the state government's commitment to harnessing AI's transformative potential while proactively mitigating associated risks. The framework is primarily composed of the Western Australian Government Artificial Intelligence Policy (Version 1) and the accompanying AI Assurance Framework, which together establish a principles-based approach to AI governance. It is designed to support government agencies in navigating the complexities of AI technologies, including both traditional AI systems and emerging generative AI capabilities, ensuring that their use aligns with public values and legislative requirements.

The core objective of this framework is to foster an enabling environment where the benefits of AI can be maximised for the community and government operations, while ensuring that all AI usage is underpinned by robust ethical considerations and risk management strategies. It mandates that AI systems are designed, built, and used transparently, safely, and reliably across their entire lifecycle. The framework encourages innovation within the public sector by providing clear guidelines and tools, such as the self-assessment mechanism, to help agencies evaluate their AI projects' implications, risks, and controls. Furthermore, it highlights the importance of human judgment and intervention in all stages of AI implementation, emphasising that AI should augment, rather than replace, human decision-making, particularly in matters significantly affecting individual rights.

Definitions

The Western Australia AI Policy Framework incorporates several key definitions to ensure a clear understanding of its scope and requirements. "Artificial Intelligence (AI)" is broadly defined as machines capable of simulating elements of human capability, encompassing a wide array of technologies from traditional AI systems to advanced generative AI. This expansive definition ensures that the framework remains relevant as AI technologies continue to evolve. The "AI Lifecycle" refers to the complete progression of an AI system, from its initial design and data modelling phases through development, validation, deployment, ongoing monitoring, and eventual refinement or decommissioning, with ethical and risk considerations being paramount at every stage.

Critical to the framework's governance structure is the concept of an "AI Accountable Officer," an executive within each government entity tasked with overseeing the ethical, safe, and responsible application of AI, including the approval of AI Assurance Framework self-assessments. The "AI Assurance Framework" itself is a systematic approach designed to facilitate the secure and reliable use of AI by guiding entities through the evaluation of AI project implications, risks, and controls. "Contestability" is a vital principle, ensuring that individuals affected by AI-driven decisions can understand the rationale behind those decisions and have access to transparent review mechanisms to challenge AI-informed outcomes. This requires sufficient insight into the algorithms and inferences made. Finally, "Privacy by Design" mandates that privacy considerations and data protection measures are embedded into AI systems from their inception, while "Generative AI" is explicitly included within the policy's scope, acknowledging the rise of AI capable of producing new content.

Governance and Institutional Framework

The governance structure for Artificial Intelligence within the Western Australian public sector is meticulously outlined in the Western Australia AI Policy Framework, with the Office of Digital Government (DGov) playing a central role under the Department of the Premier and Cabinet. A cornerstone of this framework is the mandatory requirement for all government entities to nominate an "AI Accountable Officer". This executive-level position is crucial for ensuring that AI initiatives within their respective agencies align with the overarching AI Policy and Assurance Framework. The AI Accountable Officer is responsible for the oversight of AI governance and the approval of AI Assurance Framework self-assessments, thereby embedding accountability at a senior level. This institutional arrangement aims to provide clear lines of responsibility and foster a culture of ethical AI adoption across the public sector.

Beyond the individual agency level, the framework establishes a broader governance ecosystem. The Office of Digital Government (DGov) is tasked with developing and updating the AI Policy materials and guidance, ensuring they remain contemporary as the AI landscape evolves. DGov also manages the AI Assurance Framework's self-assessment process and maintains the WA Government's AI Register, which includes submissions from agencies. Furthermore, the establishment of an Artificial Intelligence Advisory Board, created by DGov, provides strategic direction and expert insights on the responsible and ethical integration of AI technologies across government services. This multi-layered approach, involving accountable officers, a central coordinating body, and an advisory board, is designed to strengthen AI governance, promote safe innovation, and ensure public trust in the government's use of AI.

Key Focus Areas

The Western Australia AI Policy Framework is built upon a set of core principles that guide the ethical, safe, and responsible use of AI across the public sector. A primary focus is on ensuring that AI systems deliver tangible "benefit to the community and agency," requiring careful consideration of how AI solutions can improve productivity, assist stakeholders, and enhance agency capabilities. "Accountability" is another critical area, mandating clear identification of responsibility for AI systems and their outcomes throughout the entire AI lifecycle, with a particular emphasis on independent review for systems significantly impacting individual rights. This is closely linked to "transparency and disclosure," which requires that meaningful information about AI usage is made available to affected stakeholders, enabling them to understand outcomes and, if necessary, challenge them.

Furthermore, the framework places significant emphasis on "privacy protection and security," asserting that respect for privacy and data protection is paramount. This includes ensuring proper data governance, management, and the incorporation of "privacy by design" principles, alongside robust security measures to protect against vulnerabilities and adversarial attacks. "Reliability and safety" are also key, demanding that AI systems operate consistently with their intended purpose, are accurate, reproducible, and do not pose unreasonable safety risks, with mitigation measures proportionate to potential harms. Finally, "fairness and non-discrimination" are central, requiring measures to identify and manage potential data biases and ensure that AI-produced decisions comply with anti-discrimination laws, promoting equitable access and treatment for all individuals.

Implementation Framework

The implementation of the Western Australia AI Policy Framework is primarily facilitated through the mandatory AI Assurance Framework and its associated self-assessment process. This framework provides a systematic and practical approach for government entities to evaluate their AI projects, ensuring alignment with the established policy principles. Agencies are required to use the self-assessment tool for any system or project involving AI or automated decision-making, regardless of whether the AI system is procured, built, or adapted. This comprehensive application ensures that ethical and risk considerations are integrated at all phases of an AI project, from initial design and planning through development, deployment, and ongoing operation. The self-assessment guides project leads through a structured process of describing the project's benefits, assessing privacy, security, transparency, explainability, and contestability risks, and outlining appropriate risk controls.

A crucial element of the implementation framework is the approval process for these self-assessments. Once completed by the AI project lead, the self-assessment is submitted to the entity's nominated AI Accountable Officer for review and approval. This executive-level endorsement ensures that AI initiatives have appropriate oversight and are consistent with the agency's broader governance. The Office of Digital Government (DGov) also plays a supportive role, offering guidance and including approved self-assessments in the WA Government's AI Register. The framework is designed to complement, rather than duplicate, existing Information and Communication Technology (ICT) governance within entities, encouraging the use of existing project documentation where possible. This integrated approach aims to embed responsible AI practices within the public sector's operational fabric, fostering innovation while rigorously managing risks.

Monitoring and Evaluation

The Western Australia AI Policy Framework places significant emphasis on continuous monitoring and evaluation to ensure the ongoing responsible, safe, and effective use of AI systems within the public sector. A core requirement is that AI systems must be regularly monitored and tested to verify they continue to meet their intended purpose. This ongoing scrutiny is vital for identifying any emerging problems or deviations from expected performance, which must then be addressed through appropriate risk management strategies. The framework also stresses the importance of assessing the soundness of the connection between data and the inferences drawn by AI systems, ensuring that these connections remain robust and reliable over time. This proactive approach to monitoring helps to maintain the integrity and trustworthiness of AI solutions throughout their operational lifecycle.

Furthermore, the framework mandates that responsibility for ensuring an AI system's robustness and safety must be clearly and appropriately identified. This accountability extends to the continuous assessment of AI systems for accuracy, fairness, and bias, with regular reviews and audits being essential practices. For instance, the WA health system, in alignment with the broader government framework, employs an Artificial Intelligence Audit Compliance Tool. This tool requires WA health entities to annually report and submit details regarding policy and standard compliance, identified breaches, and system governance. Such mechanisms facilitate systemic oversight and enable trend analysis, allowing the Office of Digital Government to monitor overall compliance and adapt the framework as the AI landscape evolves, ensuring sustained ethical and responsible AI deployment.

Penalties, Liability, and Appeals

While the Western Australia AI Policy Framework is primarily a policy and assurance guide rather than a legislative instrument with explicit penalties, it establishes clear obligations and mechanisms for addressing non-compliance and ensuring accountability. The framework mandates that mechanisms must be in place to ensure responsibility and accountability for AI systems and their outcomes at all stages of the AI lifecycle. This includes identifying the organisation and individual accountable for decisions made by AI systems, particularly when those decisions have a significant impact on an individual's rights. In such cases, the framework requires that AI systems be accountable to independent review, with timely, accurate, and complete information provided to oversight bodies to facilitate this process.

A critical component for redress and appeals is the requirement for an "efficient and transparent review mechanism" accessible to the people of Western Australia. This mechanism is intended for situations where there are questions about the use of data or AI-informed outcomes. It necessitates disclosing when an AI system is engaging with individuals or when AI has been used to produce an outcome that impacts them. Furthermore, there must be sufficient access to and understanding of the algorithm and the inferences drawn to make contestability effective. This focus on transparency and the right to challenge ensures that individuals have avenues for recourse if they believe an AI system has made an unsafe, unfair, or biased decision, thereby upholding fundamental rights and fostering public trust.

Relationship to Other Instruments

The Western Australia AI Policy Framework is designed to operate within a broader legislative and policy landscape, demonstrating a commitment to coherence and integration with existing regulatory instruments. A key relationship exists with the WA Privacy & Responsible Information Sharing (PRIS) Act 2024 and the Commonwealth Privacy Act 1988. The framework explicitly states that data used in AI systems must be handled safely and in accordance with relevant legislation, underscoring that respect for privacy and data protection is paramount. This includes ensuring proper data governance and management throughout the AI lifecycle, with an emphasis on principles such as data anonymisation and informed consent, and the incorporation of "privacy by design" principles.

Furthermore, the framework aligns with and references other critical government policies and standards. It requires compliance with the WA Government Information Classification Policy for data classification and adherence to the WA Government Cyber Security Policy for security measures, including the identification of potential security vulnerabilities and assurance of resilience to adversarial attacks. The framework also draws upon the National Framework for the assurance of AI in government, ensuring that Western Australia's approach is consistent with broader national guidelines for responsible AI use. This interconnectedness with various legal and policy instruments ensures a holistic and robust regulatory environment for AI within the Western Australian public sector, promoting responsible innovation while safeguarding public interests.

International Alignment

The Western Australia AI Policy Framework demonstrates a clear commitment to aligning with established international best practices and ethical principles for Artificial Intelligence. The framework explicitly states its adherence to the OECD AI Principles (2019), which advocate for the innovative and trustworthy use of AI while respecting human rights and democratic values. This alignment ensures that Western Australia's approach to AI governance is consistent with globally recognised standards for responsible AI development and deployment. The OECD principles, which promote AI for beneficial outcomes, respect for the rule of law, human rights, freedom, dignity, autonomy, privacy, data protection, non-discrimination, equality, diversity, fairness, and transparency, form a foundational ethical compass for the WA framework.

Moreover, the framework is developed with an awareness of and aims to align with the broader national and international AI landscape. It works collaboratively with other Australian states and territories to ensure best practice is leveraged and that there is alignment across jurisdictions, particularly as regulatory approaches evolve at both Commonwealth and state government levels. This includes drawing upon content developed by the NSW Government and the Australian Government Department of Industry, Science and Resources, indicating a harmonised approach within Australia. The emphasis on human-centred values, fairness, privacy protection, reliability, transparency, contestability, and accountability within the WA framework directly mirrors the Australian Government's AI Ethics Principles, further solidifying its international alignment and commitment to a globally responsible AI ecosystem.

Implementation Timeline

Milestone	Date	Notes
WA Government AI Policy and Assurance Framework Enactment	2024-03-05	Official release date for the foundational policy and assurance framework documents.
Nomination of AI Accountable Officers Deadline	2025-09-19	Government entities must nominate an executive to oversee AI use and approve self-assessments by this date.
Ongoing Policy Updates	As required	The Policy will be updated as AI technologies evolve and understanding of benefits and risks develops, ensuring its continued relevance.
Regular Self-Assessment Application	Throughout AI Project Lifecycle	Entities must use the AI Assurance Framework self-assessment at all phases of AI projects (design, development, deployment, monitoring) and for existing AI solutions.
AI Advisory Board Establishment	2024-11-07	Expressions of Interest launched for the newly formed AI Advisory Board to provide strategic direction and insights.

Compliance Checklist

Check	Required Action
Nominate AI Accountable Officer	Each government entity must nominate an executive with responsibility for overseeing AI use and approving AI self-assessments.
Complete AI Assurance Framework Self-Assessment	Conduct self-assessments for all AI systems or projects, including procured, built, or adapted solutions, at all phases of the project lifecycle.
Ensure Human Oversight and Intervention	Implement mechanisms to ensure human judgment and intervention are vital in harnessing AI's potential, especially for decisions significantly affecting rights.
Adhere to Ethical Principles	Apply the established AI Ethics Principles (e.g., benefit to community, accountability, transparency, fairness, privacy, security, reliability, safety, contestability) in all AI usage.
Comply with Data Protection Laws	Use data safely and in accordance with the WA Privacy & Responsible Information Sharing Act 2024 and the Commonwealth Privacy Act 1988, incorporating privacy by design.
Implement Robust Security Measures	Ensure appropriate data and AI system security measures are in place, including identification of vulnerabilities and resilience to adversarial attacks.
Monitor and Test AI Systems	Regularly monitor and test AI systems to ensure they continue to meet their intended purpose, addressing any identified problems with ongoing risk management.
Provide Transparency and Explainability	Make meaningful and transparent information available on how AI is being used, enabling affected stakeholders to understand outcomes and challenge them.
Establish Review Mechanisms	Ensure access to an efficient and transparent review mechanism for questions about data use or AI-informed outcomes, with sufficient understanding of algorithms for contestability.
Manage Data Bias and Quality	Implement measures to identify and appropriately manage potential data biases and ensure data quality and relevance for AI systems.

Sources and References

Source	Type
Artificial Intelligence Policy - Western Australian Government	official
Artificial Intelligence \| Office of Digital Government - Department of the Premier and Cabinet	government

The Western Australia AI Policy Framework sets out how the state's public sector must ethically and responsibly develop and use Artificial Intelligence, from traditional systems to emerging generative AI.

This policy, which came into effect on March 5, 2024, applies to all Western Australian government agencies and entities. It mandates a structured approach to AI, requiring them to: - Nominate an "AI Accountable Officer" by September 19, 2025. This executive will oversee AI use and approve project assessments. - Complete an "AI Assurance Framework self-assessment" for every AI project, whether built, bought, or adapted, at all stages of its lifecycle. These assessments evaluate risks related to privacy, security, transparency, and fairness. - Ensure human oversight and intervention, especially for decisions significantly affecting individual rights. AI should augment, not replace, human judgment. - Embed "Privacy by Design" and robust security measures from the outset, complying with WA and Commonwealth privacy laws.

While not a law with direct penalties, the framework demands clear accountability for AI systems and their outcomes. Crucially, it requires transparent review mechanisms, allowing individuals affected by AI-driven decisions to understand the rationale and challenge them. This "contestability" means agencies must provide sufficient insight into algorithms and inferences. A key practical takeaway is that the policy covers *all* AI use, including off-the-shelf generative AI tools, meaning every instance needs careful assessment and oversight to ensure ethical and safe deployment. The Office of Digital Government oversees compliance and maintains an AI Register.

Australia - Western Australia - AI Policy Framework (RAI-AU-WA-WESAUAI-2024)

Summary

Plain English