APS Tests Generative AI for Government Use Safety

The trial runs from January to June 2024, coordinated by the Digital Transformation Agency (DTA). It allows APS staff and agencies to innovate and explore how generative AI might support routine, day-to-day tasks - such as meeting summaries and action items, internal correspondence, or presentation outlines - to free more of the day for detailed, substantive work that people do best.

Chris Fechner, Chief Executive Officer for the DTA, says that the government is committed to embracing AI in a safe, ethical and responsive way, and fostering an innovative culture in the APS.

"Government should be an exemplar in transparency, risk management and governance of emerging technologies such as generative AI".

"By having agencies with different types of work participate - from the Department of Industry, Science and Resources to the Department of Finance and the National Disability Insurance Agency - the Australian Government will better understand how to harness generative AI in a safe, ethical and responsible way."

A complete list of participating agencies is available below.

Guidance for participating agencies

"The DTA, in partnership with the AI in Government Taskforce, is leading the evaluation of the trial," Mr Fechner explains.

Participating agencies are rolling out the generative AI tool according to their own circumstances and requirements, including their security and technical readiness.

Participating staff are required to undertake learning module and knowledge assessment specific to the trial. Both of the learning tools build upon the existing Interim guidance for agency use of generative AI that offer direction through the application of golden rules, tactical guidance, use cases, and five principles in practice:

1. Accountability. APS staff-members must be able to explain, justify, and take ownership of any advice or decisions that where generative AI tools were used to assist the process.
2. Transparency and explainability. Agencies should consider appropriately marking where generative AI was used, and any information generated must be critically examined.
3. Privacy protection and security. The government should not use public generative AI tools with any classified, personal, or sensitive information and should follow the relevant laws and policies.
4. Fairness and human-centred values. The government should avoid biases in generative AI tools that can harm some groups and should involve relevant communities in decision-making.
5. Human, societal, and environmental wellbeing. The government should use generative AI tools in a way that improves the wellbeing of the community, respects right holders, as well as carefully considering Indigenous data sovereignty and governance.

These five principles in practice expand on Australia's 8 AI Ethics Principles, a voluntary framework to reduce the risk of negative impact by maintaining safe, reliable, and fair outcomes for Australians.

Australia's data and digital approach

This trial continues the Australian Government's commitment to improve how the APS adopts and uses data and digital technologies as outlined by the Digital and Data Government Strategy (the Strategy). This strategy identifies several areas of improvement, including enhancing service delivery, advancing interoperability, replacing outdated technologies, exploring the use of new technologies, updating legislation, and leading by example.

Mr Fechner links this latest generative AI trial, as driven by the Strategy, to Australia's recent recognition as a world leader in digital government. Australia debuted at 5th in the Organisation for Economic Co-operation and Development's (OECD) Digital Government Index for 2023.

"The APS and the DTA will keep looking for opportunities to lift our Digital Government performance as evidenced by our inaugural OECD ranking and strive to improve on it in future years for the benefit of all Australians," says Mr Fechner.

"These goals go hand in hand with the Government for the Future mission that strives to take advantage of emerging technologies to deliver secure, ethical, and modern data and digital technologies by 2030."