The Creation and Analysis of Government AI Transparency Statements in Australia

Transparency requirements exist but implementation varies widely: Australia's AI Transparency Statements standard requires government agencies to disclose their AI use, but the analysis of 101 actual statements (AITS-101 dataset) reveals "substantial variation in AI-related practice disclosure," meaning some agencies provide much more detailed information than others about how they're using AI in public services.

Gap between policy goals and real-world practice: The paper identifies a significant disconnect between what the transparency policy is designed to achieve and what actually appears in government documents, suggesting that having a transparency standard on paper doesn't automatically guarantee meaningful public disclosure of AI use.

Disclosure gaps undermine accountability: The findings highlight specific gaps in what government agencies are disclosing about their AI systems, which matters because transparency is framed as essential for both "accountability and public trust"—without complete information, the public cannot effectively evaluate whether AI is being used responsibly in services that affect them.

Document structure and quality affect usefulness: Through analysis of how statements are written and organized, the paper reveals that the way transparency information is presented matters for whether the public can actually understand and evaluate government AI use, suggesting that clearer standards for how to disclose are needed, not just that disclosure must occur.

Findings can improve future standards: By systematically analyzing what works and what doesn't in current transparency statements, this research provides evidence-based recommendations for designing more effective AI transparency requirements for government.