Submission 14
Centre for Public Integrity
Submitter group: Research institutes.
An AI model identified the submitter group and wrote the summary, key points and recommendation summaries. Counts do not measure importance or merit.
Summary
Read the original submission for the full text.
Argues for systemic information-integrity policy supporting democratic institutions, and a Commonwealth automated decision-making framework covering transparency, oversight, meaningful human judgment and review rights. Says safeguards should vary with the technology, its role and the consequences of decisions, rather than impose identical requirements on all systems.
Key points
Quotes are the submitter’s own words. Text taken from documents, including scanned pages read by text-recognition software, can differ from the PDF. Check the original. Listed in the same order as on the submission’s page in the interactive guide.
Argues that information integrity should be understood as a foundational democratic principle.
The Centre’s developing work in this area argues that information integrity should be understood as a foundational democratic principle.
Argues that aI-generated misinformation should not be treated only as a new harmful-content category; the response should be systemic and institutional.
The Centre cautions, however, against treating AI-generated misinformation simply as a new category of harmful content requiring a discrete regulatory response. The challenge is broader and systemic.
The Centre is concerned that increasing use of ADM may create significant accountability gaps unless existing public law frameworks are adapted.
The Centre is concerned that the increasing use of ADM may create significant accountability gaps unless existing public law frameworks are adapted accordingly.
Argues that a human in the loop does not resolve these problems unless oversight is genuine and meaningful.
The formal retention of a “human in the loop” will not necessarily resolve these problems. Where an automated system informs or recommends an administrative decision, human oversight must be genuine and meaningful: the responsible officer must have sufficient information and capacity to scrutinise the automated output, rather than simply endorsing it.
Argues that procurement should not allow governments to contract out of transparency through commercial-in-confidence claims.
Government should not be permitted to contract out of transparency and accountability obligations by relying upon commercial-in-confidence claims by private technology providers.
Recommendations
Recommendations are shown beside the submission’s own words. Listed in the same order as on its page in the interactive guide.
Maintain strong integrity frameworks that limit undue influence, enhance executive accountability and sustain public trust.
Governments must maintain strong integrity frameworks that limit undue influence, enhance executive accountability and sustain public trust.
Protect and support independent knowledge institutions such as the AEC, CSIRO, Bureau of Meteorology, universities, courts and expert regulators, including their independence and adequate resourcing.
Governments must protect and support independent ‘knowledge institutions’. Institutions such as the Australian Electoral Commission, CSIRO, Bureau of Meteorology, universities, courts, expert regulators and other independent public bodies play a critical role in producing, testing and disseminating reliable knowledge. Their independence, expertise, adequate resourcing and public credibility provide essential infrastructure for an information environment increasingly populated by synthetic and potentially unreliable material.
Provide proactive, accessible, authoritative information and responses to significant misinformation within institutional expertise, with transparent, evidence-based public communication insulated from partisan manipulation.
Governments themselves have a responsibility to provide clear, accessible and authoritative public information. Information vacuums create opportunities for misleading or fabricated material to take hold. Public institutions should therefore proactively make reliable information available and be equipped to respond to significant misinformation concerning matters within their expertise. The credibility of this function depends upon public communication being transparent, evidence- based and appropriately insulated from partisan manipulation.
Invest in media, digital and AI literacy as core democratic competencies, through long-term initiatives across education systems and public institutions.
Governments must invest in media, digital and AI literacy. Citizens increasingly need the capacity not only to assess the reliability of sources and recognise manipulation, but to understand how algorithmic systems curate information and the possibilities and limitations of generative AI. These capabilities should be treated as core democratic competencies and supported through long-term initiatives across education systems and public institutions.
Support legitimate deliberative forums that enable informed disagreement.
Governments should support legitimate deliberative forums that enable informed disagreement.
Widen access to decision-making and public debate.
Governments must widen access to decision-making and public debate.
Support a robust and independent public-interest media ecosystem.
Governments must support a robust and independent public-interest media ecosystem.
Approach AI’s information-integrity implications systemically, maintaining the institutional conditions for a healthy democratic information environment.
The Centre therefore encourages the Committee to approach the implications of AI for information integrity systemically. The central policy question is not only how Australia should regulate false or misleading AI-generated content, but how the state can maintain the institutional conditions necessary for a healthy democratic information environment as AI changes the way information is produced, distributed and consumed.
Begin the regulatory framework by distinguishing AI and automated decision-making technologies and uses, including algorithmic decision-making and emerging generative AI.
A regulatory framework should begin by recognising that “AI” and “automated decision- making” encompass different technologies and uses. There is, for example, an important distinction between algorithmic decision-making, involving rules or models that process information and produce or inform outcomes, and the emerging use of generative AI, including large language models.
Match safeguards to the technology used and its significance in the exercise of public power.
These distinctions matter. The appropriate safeguards should reflect both the technology being used and the significance of its role in the exercise of public power.
Adopt a comprehensive framework for the transparent and accountable use of ADM across Commonwealth government.
the Centre recommends that the government adopt a comprehensive framework for the transparent and accountable use of ADM across Commonwealth government.
Conduct an initial retrospective audit of government ADM, followed by an ongoing mechanism to identify and record its use.
There should be an initial retrospective audit, followed by an ongoing mechanism for identifying and recording the use of ADM across Commonwealth departments and agencies.
Disclose where and when ADM is used and meaningful, accessible system information, including relevant algorithms, code, datasets, models and prompts; consider a central public register or repository.
A strong presumption of transparency. Government should disclose where and when ADM is being used and provide meaningful and accessible information about the systems involved, including, where relevant, algorithms, code, datasets, models and prompts. Consideration should be given to establishing a central public register or repository for this information.
Adopt procurement rules so government cannot contract out of transparency and accountability obligations through commercial-in-confidence claims, and retains the rights and information needed to explain, scrutinise and review systems.
Government should not be permitted to contract out of transparency and accountability obligations by relying upon commercial-in-confidence claims by private technology providers. Commonwealth procurement arrangements should ensure that government retains the rights and information necessary to explain, scrutinise and review systems used in the exercise of public power.
Consider independent ADM oversight to maintain a register, coordinate audits, develop expertise and guidance, support agency compliance and help individuals understand ADM’s role in decisions affecting them.
Consideration should be given to an independent office or other institutional mechanism responsible for overseeing government use of ADM. Its functions could include maintaining the register or repository; conducting or coordinating audits; developing expertise and guidance; supporting agencies to comply with their obligations; and providing a point of contact for individuals seeking to understand whether and how ADM has been used in a decision affecting them.
Require legal and integrity review of proposed ADM uses before deployment, particularly where they materially contribute to decisions affecting individuals, possibly including mandatory consultation with the Attorney-General's Department and consultation with the oversight body.
Proposed government uses of ADM, particularly where they will materially contribute to statutory or administrative decisions affecting individuals, should be subject to legal and integrity review before deployment. This could include mandatory consultation with the Attorney-General’s Department regarding the legality of the proposed deployment of ADM, and consultation with the proposed specialist oversight body.
Assess ADM systems before and during deployment for bias, discrimination and systemic unfairness, with independent scrutiny and public reporting.
ADM systems should be assessed before (for example, through a trial) and during deployment for bias, discrimination and other forms of systemic unfairness, with appropriate independent scrutiny and public reporting.
Ensure that where legislation requires human judgment or discretion, the responsible decision-maker genuinely exercises it rather than endorsing an automated outcome.
Where legislation requires the exercise of human judgment or discretion, administrative arrangements must ensure that the responsible decision-maker genuinely exercises that power rather than simply endorsing an automated outcome.
Ensure ADM does not diminish existing rights to reasons, merits review and judicial review, and that individuals can access enough information to challenge ADM decisions.
Australia’s administrative law framework should ensure that the use of ADM does not diminish existing rights to reasons, merits review and judicial review. Individuals must have access to sufficient information about the operation of ADM to understand and meaningfully challenge decisions affecting them.
Require genuine human scrutiny of automated administrative recommendations, with officers having sufficient information and capacity to scrutinise outputs.
Where an automated system informs or recommends an administrative decision, human oversight must be genuine and meaningful: the responsible officer must have sufficient information and capacity to scrutinise the automated output, rather than simply endorsing it.
Make ADM requirements proportionate to the technology, its role and decision consequences, preserving transparency, accountability, fairness and reviewability.
The Centre does not suggest that all automated systems should be subject to identical requirements. The framework should be proportionate to the nature of the technology, the role it performs, and the consequences of the decision. But the underlying principle should be clear: the adoption of new technology by government must not diminish the transparency, accountability, fairness or reviewability of public power.