APS Issues Official Guidelines on Use of Generative AI in the Public Sector
The Digital Transformation Agency (DTA) has issued interim guidance on the use of Generative Artificial Intelligence (AI) in the public sector, marking an important development for the Australian Public Service (APS). The guidelines, which are designed to establish guardrails for the use of AI platforms such as ChatGPT, Bard AI, and Bing AI, highlight the need for public servants and suppliers to record and permission the use of generative AI to ensure proper oversight.
The DTA advises agencies to implement an enrolment mechanism for staff user accounts, with approval processes overseen by Chief Information Security Officers (CISO) and/or Chief Information Officers (CIO). This step aims to address potential unauthorized or unregulated use of generative AI within government departments. Moreover, agencies are required to self-report any breaches of the AI policy, allowing for transparency and accountability.
While AI has been utilized within the government for years, the DTA and Department of Finance are primarily concerned about the use of publicly available generative AI platforms. These platforms, which may not undergo security risk assessments or be covered by commercial contracts with agencies, pose a potential risk of introducing biases into clean or trusted datasets and models.
Collaborating with the Department of Industry, Science, and Resources, the DTA has crafted these guidelines to ensure the ethical use of AI in government. The principles aim to promote responsible, safe, and ethical practices surrounding the use of AI by the APS. The guidelines emphasize that sensitive or classified information should not be uploaded or input into public AI platforms. Instead, non-sensitive information from government websites or public agency reports can be used or generic slide templates can be requested to create presentations.
The guidelines also address potential pitfalls, such as uploading tender information. Public servants are advised to exercise caution when using AI platforms to search for technical information and avoid mentioning sensitive details or organizational information that could indicate the preparation of a tender. It is recommended that the information obtained from AI platforms be validated by a human for accuracy and appropriateness before use.
The DTA acknowledges that the guidance provided will evolve over time, and encourages public servants to follow their agency’s policies and guidance on the use of generative AI tools. This iterative approach ensures that agencies can adapt their practices as needed while adhering to the overarching principles set forth by the DTA.
In conclusion, the issuance of official guidelines on the use of generative AI in the public sector represents a significant step towards ensuring responsible and ethical AI practices within Australian government agencies. With the implementation of these guidelines, the APS aims to harness the benefits of AI technology while mitigating potential risks and safeguarding data integrity.
