The Australian government has revealed ambitious plans to embed artificial intelligence across all public service departments, despite acknowledging significant security concerns and data risks identified during recent trials.
Widespread AI Adoption Planned
Finance Minister Katy Gallagher announced the government's intention to see AI 'widely taken up across government in every department' following a six-month trial of Microsoft Copilot. The initiative forms part of a broader strategy to harness the productivity benefits of generative AI programs like ChatGPT and Gemini.
Gallagher, who also serves as minister for the public service, told a conference that the government aims to 'take hold of the opportunities that AI presents' while addressing the challenges that emerged during testing.
Trial Results: Promising but Problematic
The government's comprehensive trial of Microsoft Copilot yielded mostly positive feedback, with surveys showing that 69% of participants reported working faster and 61% noted improved work quality. Management and executive staff particularly benefited, saving up to an hour daily through AI-assisted tasks like information summarisation, drafting documents, and research.
However, the trial also uncovered significant issues that require attention before full implementation:
- 60% of users needed to make moderate to significant edits to AI-generated content due to inaccuracies
- Concerns about Copilot's unpredictability and lack of contextual knowledge
- Instances of the system accessing sensitive information inappropriately
- Potential disproportionate impact on female administration staff
Security Concerns and Data Risks
The trial report highlighted serious security implications, noting that Copilot enabled some participants to access documents they shouldn't have had permission to view. This occurred primarily because organisations hadn't properly secured data storage before adopting the AI tool.
The report warned that 'without the appropriate data infrastructure and governance in place, the use of Copilot may further exacerbate risks of data and security breaches in the APS'.
Public scrutiny remains high following the Robodebt scandal, with particular attention on how the public service uses technology for automated decision-making.
Future Plans and Union Consultation
The government is developing its own AI program called GovAI Chat, expected to roll out widely in early 2026. The AI plan sets a goal that 'every public servant will have training and access to generative AI tools'.
Gallagher stressed that the government 'does not view widespread AI adoption across the APS as a way of replacing people' and pledged ongoing consultation with staff and unions about job changes. The Community and Public Sector Union has been contacted for comment on the proposed changes.
Microsoft's general manager of public sector, Vivek Puthucode, endorsed the government's approach, stating the technology could 'help improve service delivery, strengthen policy outcomes and make government more responsive to community needs'.
The government will also develop guidance for public servants using public AI platforms to handle government information classified up to 'official' level, balancing innovation with necessary security precautions.
