Back to Articles
Senate QON 542 - Senator Jacqui Lambie, Social Services
Parliament of Australia
ENRICHED
Details
- Date Published
- 21 May 2026
- Priority Score
- Not scored
- Australian
- Yes
- Created
- 16 Sept 2026, 09:00 pm
Authors (0)
No authors linked
Description
Could the Minister please advise on the following: What artificial intelligence [AI] system/s are being used to assess submissions from the public and organisations and in relation to these: please detail all systems in use including the architecture/technique and functional role of the AI systems in use, for example large language models or Generative AI; and where is the data/information used by the AI systems stored, is that information held in servers overseas. What has the Department done t
Body
QUESTION
Could the Minister please advise on the following: What artificial intelligence [AI] system/s are being used to assess submissions from the public and organisations and in relation to these: please detail all systems in use including the architecture/technique and functional role of the AI systems in use, for example large language models or Generative AI; and where is the data/information used by the AI systems stored, is that information held in servers overseas. What has the Department done to inform Australian citizens that their submissions are being assessed by AI. Are AI models being used to assess applications for government support programs (for example the National Rental Affordability Scheme/Home Equity Access scheme, Youth Allowance, abstudy, etcetera); please detail all. Is AI being used to assist, adjust or alter any of the above income payments. What checks are in place to ensure the claim is being assessed appropriately. Can the Department define what type of AI systems are in use (for example algorithmic systems or advanced automation). Do these systems have human oversight, and if so: how many people; and please detail their qualifications and experience. Does the Department coordinate with other relevant departments on the deployment of AI. How often does the department run quality assurance checks on these systems. What checks and balances are in place to ensure that the AI assessment model employed by the department tools is operating fairly and effectively.
ANSWER
Minister Plibersek - The answer to the honourable Senator's question is as follows: The department does not use AI systems to autonomously assess or decide on public or organisational submissions. Staff may use Microsoft Copilot Chat as a productivity tool to assist with activities such as summarising or analysing submissions as part of their normal work. Microsoft Copilot Chat (Microsoft 365) - large language model based generative AI. All departmental information is stored onshore in Australian Government approved data centres. AI is not used to autonomously assess or decide on public submissions. No. AI is not used to autonomously assess or decide on applications for government support programs that the department administers. No. AI is not used to calculate, determine, adjust or alter payments for government support programs that the department administers. All assessments are undertaken by human decision-makers in accordance with legislation, policy, and established assurance processes. The department uses: Generative AI and algorithmic systems for internal productivity Low-risk automation is used to handle some simple rules-based tasks for internal productivity. Yes. All AI outputs are reviewed by departmental staff prior to use Oversight is provided by departmental staff with relevant policy, legal, program, JCT, and governance expertise. Yes, the department aligns AI use with: Digital Transformation Agency policy Whole-of-government AI standards Department of Finance Gov AI platform Australian Public Service Commission capability and ethics frameworks. AI use is subject to: Use-case approval and review Monitoring of usage patterns Ongoing governance oversight Periodic reviews aligned to risk and impact. Controls include: Approved use-case restrictions Prohibition on automated decision-making Human-in-the-loop review of all outputs Privacy, legal, security and records compliance requirements.