Recommendations
GovAI Chat
Universally accessible and secure chat tool
Lead agency: Finance
The government will provide access to secure generative AI, through GovAI Chat, for everyone in the Australian Public Service. GovAI Chat will provide the capabilities of modern generative AI tools, while allowing users to leverage government data. Responses will be fast, current, and auditable through familiar and secure interfaces. Where smaller agencies may lack the scale and resourcing to adopt native AI tools, this will support staff across all agencies to safely and effectively experiment with AI tools and integrate them within their workflows.
The underlying framework will be flexible to respond to user needs and feature requests as technology evolves. It will be rolled out in an iterative fashion to allow for extensive and ongoing user testing with users and teams, across roles and agencies to ensure it meets practical needs.
This will increase the capability of the APS to deliver more for the Australian public, while maintaining standards and complying with legal frameworks for cyber security and data sovereignty. The result will be a safer and smarter way to work that supports operational integrity and delivers value through time saved and reduced rework.
Guidance on public and enterprise AI services
Clarity and consistency on using public LLMs up to OFFICIAL
Lead agency: DTA/Home Affairs
The government will develop clear guidance on the use of public generative AI tools to give public servants confidence to use platforms like ChatGPT, Claude, and Gemini up to OFFICIAL level information. The Department of Home Affairs will issue policy guidance to outline requirements for using these tools via web browsers. DTA will also update its existing guidance on the use of public generative AI to reflect this and include practical examples to support responsible use.
The work will help shift the risk appetite across agencies and enable staff to access general-purpose AI tools at no cost. Each agency will need to implement the changes independently and should consider using safeguards like upload blockers to manage risks around classified or personal information or other sensitive information.
Support for AI tool procurement
Speeding processes, sharing information, and assisting agencies
Lead agency: DTA
The government will continue to enhance AI procurement pathways to make it easier for agencies to access trusted AI products and services in line with government standards. This will help address concerns from agencies that are cautious about adopting tools when changes like added generative AI features are introduced without prior notice or clear guidance. DTA is introducing AI-specific subcategories within procurement panels like BuyICT and Digital Marketplace, helping agencies identify vendors with proven capabilities and adopt AI tools consistently and quickly. This also helps ensure alignment with the government’s technical and ethical standards, reducing duplication and accelerating the safe adoption of AI solutions across the APS.
The DTA is also developing guidance to support AI procurement across government. The guidance will include a practical procurement checklist and aims to help agencies find and manage AI-related risks while supporting procurement best practices. The Department of Finance and the DTA are working together to ensure that current procurement arrangements remain fit-for-purpose for a rapidly evolving supplier ecosystem and new business offerings.
Women are significantly underrepresented across STEM and mixed STEM occupations. As government plays a key market-shaping role, future consideration will be given to better integrating gender equity objectives in AI or STEM-related procurement to ensure public investment aligns with both productivity and social equity goals, which will help deliver on objectives of Working for Women: A Strategy for Gender Equality.
Additionally, procurement processes, risk, and security assessments are often duplicated across agencies due to low visibility. AI adoption is slowed and disincentivised where agencies are faced with these processes.
To better serve the public, every public servant will have:
- foundational training and capability supports to use generative AI tools safely, responsibly and effectively
- access to generative AI tools
- clear guidance on how to use these tools responsibly
- leadership and support from Chief AI Officers to promote adoption
- opportunities to collaborate, build on and reuse work of others.
What we plan to achieve
The AI Plan for the APS positions the government to improve service delivery, policy outcomes, efficiency and productivity by substantially increasing the safe and responsible use of AI in government.
Strategic alignment
Data and Digital Government Strategy
This plan supports progress toward the government’s 2030 vision to deliver simple, secure and connected public services, for all people and businesses, through world-class data and digital capabilities.
Off
APS Reform Agenda
This plan aligns with and builds on the APS Reform agenda by embedding digital skills in the workforce and driving sustainable AI adoption to build a stronger APS that delivers better outcomes for the community.
OffThese targets are based on an 18-month timeframe. Broadly, these align with two overarching milestones:
- July 2026 (0-12 months)
- July 2026 onwards (12+ months).
Individual initiatives also contain their own targets and measures (see Appendix A).
As an iterative plan, targets and delivery schedules will be updated to account for new developments and emerging opportunities.
- Everyone in the APS completing training on the fundamentals of AI use in the APS and having access to guidance to use AI safely and securely
- All public servants having access to generative AI tools
- Each agency and department appointing a senior executive as Chief AI Officer
- All agencies tracking and reporting their AI use
APS AI Plan implementation timeline
Image description
APS AI Plan timeline from July 2025 to December 2026 showing milestones for the 15 initiatives across the Trust, People, and Tools pillars.
Under Trust: AI in government policy and guidance updates (July to December 2025), AI Review Committee (December 2025 to December 2026), Clear expectations of external service providers (December 2025 to December 2026), and AI strategic communications (July 2025 - December 2026).
Under People: Foundational learning (July 2025 - December 2026), Staff consultation and engagement (July 2025 - December 2026), AI delivery and enablement (December 2025 to December 2026), and Chief AI Officers (Dec 2025 to December 2026).
Under Tools: GovAI open trial (July to October 2025), GovAI: Centrally hosted AI services (November 2025 to December 2026), GovAI Chat (December 2025 to December 2026), Guidance on public and enterprise AI services (July to December 2025), Support for AI tool procurement (December 2025 to July 2026), Re-using intellectual property (December 2025 to December 2026), Central register of generative AI assessments (December 2025 to July 2026), and New whole-of-government cloud policy (November 2025 to March 2026).
OffAustralia's AI ambition
To strengthen Australia’s economy, society and security, the Government’s vision for AI in Australia focuses on:
- capturing the opportunities of AI
- ensuring the benefits are shared widely, and
- keeping Australians safe.
Many AI initiatives are already underway, with progress being made against these objectives. However, to date, the adoption of generative AI across government has been inconsistent. There are varying levels of AI maturity between – and even within – agencies.
AI maturity in the APS is a journey: starting with no formal adoption, progressing through leadership engagement and foundational capability, advancing to data-driven improvements in services, and advancing to AI becoming standard practice throughout government.
Agencies are also facing a complex mix of uncertainties and risks in dealing with generative AI. There are questions around the appropriate use of AI, and the advent of the latest AI developments and the speed of its advancements has heightened the need to manage privacy, cyber security and sovereignty risks. The rapidly changing digital environment will also likely introduce new risks and considerations. The plan seeks to get the balance right – capturing the opportunities and maximising the benefits, while minimising harms and mitigating risks.
Governance, training and communications initiatives already delivered
- Technical standard for government’s use of AI
- Guidance on the use of OFFICIAL information in public generative AI tools
- Sponsored GovHack 2025
- AI in government fundamentals eLearning (available for all staff)
- AI Government Showcase of AI use-cases with industry & academia
- GovAI applied eLearning
- APS Academy MasterCraft webinars for new users
- APS Academy Lunch and Learn webinars on AI use cases in the APS and how to implement AI in APS organisations delivered in partnership with GovAI
Governance, training and communications initiatives already delivered
- Let’s also get our communication right and tell our colleagues and peers what an icon is and, just as important, what it’s not and how often to use them.
- I borrow a similar armour or weapon rarity scale from your favourite RPG or shooter game.
- Another good rule to follow is that of the content it sits next to, a 16x16px icon will be lost in a hero image, large copy or large spaces. Likewise, a detailed illustration will overshadow small and short copy.
Going forward: Safe and secure data for Australians
As the government adopts and uses more AI, maintaining the security and safety of Australians’ data will be critical. Consistent with existing information security and data protection frameworks and practices, the government will take a proactive approach throughout AI implementation to give Australians confidence their data is protected.
Domestic violence survivor-victims
Provide visibility of who has access
Consider providing users with a clear and easily accessible list of who can access the service and who will be notified of any changes (e.g.
change of address). Give users the choice of when and how they receive government communications and make it easy to change, in the event they need to do it quickly.
Make it easy to remove multiple users
Support survivor-victims to remove multiple users from accessing a shared government service or account. Consider how a user can do this
in a privacy enhancing way, so as not to unnecessarily trigger or notify other users.
Support ‘quick exit’
Consider the use of ‘quick exit’ buttons within your digital service to help re-direct users to other digital pages if they are in an unsafe environment.
Clearly communicate tasks and actions
Only request information that is legislatively required and avoid unnecessary Use simple steps and actions to clearly communicate what is required and limit the impacts on survivor-victims. Consider the use of checklists and easy to follow formats to avoid decision fatigue and to support the survivor-victims to complete the service.
Off