The State of AI Transparency in Australia: A Deep Dive
A year ago, the Commonwealth government set in motion a significant policy aimed at transforming how federal agencies communicate their use of artificial intelligence (AI). By February 2025, most federal agencies are required to publish "AI transparency statements" on their websites. These statements were designed to shed light on how agencies deploy AI, the specific domains involved, and the safeguards in place. The overarching goal? To build public trust in government AI applications without relying on legislation.
However, preliminary findings from our research—set to be published in full later this year—indicate that this policy is falling short of its intentions.
The Compliance Gap
In our examination of 224 agencies, we discovered a concerning reality: only 29 agencies had easily identifiable AI transparency statements. Delving deeper, we uncovered 101 links to such statements, culminating in a compliance rate of approximately 45%. It’s worth noting that for certain agencies, like those in defence or intelligence, publishing a statement is merely recommended rather than required. Furthermore, there’s a possibility that multiple agencies could share a single statement. Nonetheless, these early findings raise alarms about the efficiency of Australia’s “soft-touch” approach to AI governance in the public sector.
The Importance of Transparency in AI Use
Public trust in AI within Australia is already teetering on the edge. The reluctance of the Commonwealth to legislate pertinent rules and safeguards regarding automated decision-making—an issue highlighted by the Robodebt royal commission—makes the demand for transparency all the more essential. Citizens expect their government to exemplify responsible AI use, yet many agencies seem to disregard the very policy intended to promote that transparency.
Beyond public discourse, good practice in government AI could inspire the private sector, which currently grapples with its own challenges regarding responsible AI. A recent study revealed that while 78% of corporations are aware of responsible AI practices, only 29% have implemented them.
More Read
Understanding AI Transparency Statements
The transparency statement is a central feature of the Digital Transformation Agency’s policy aimed at ensuring responsible AI deployment. In addition, agencies are mandated to appoint an “accountable AI official” tasked with overseeing AI utilization. The essence of these transparency statements is to be clear, consistent, and easily accessible—ideally linked right from the homepage of each agency.
Our research aimed to identify and evaluate these statements through automated website surveys, Google searches, and manual inspection of federal entity lists provided by the Office of the Australian Information Commissioner. Unfortunately, we found the majority of statements buried deep within subdomains or requiring intricate manual searches to uncover. Particularly alarming were the agencies for which publishing a statement was required, yet we still struggled to locate any.
A Toothless Requirement?
While transparency statements are theoretically binding, they prove a mute obligation in practice. The absence of penalties for agencies failing to comply severely undermines the policy’s effectiveness. Additionally, there’s no centralized system to track which agencies have published their statements, leading to a fragmented and inconsistent environment that contravenes the trust the policy aims to foster. With no clear way for the public to understand or challenge how AI is used in their lives, this lack of transparency becomes more troubling.
Global Perspectives on AI Transparency
Looking at international counterparts, we can see varying approaches to AI transparency. The United Kingdom established a mandatory AI register; however, a report by the Guardian in late 2024 suggested that many departments still neglected to list their AI systems, despite the legal requirements. Although slight improvements have been noted, high-risk AI applications identified by civil society organizations remain absent from the UK government’s register.
In contrast, the United States has taken firmer steps toward ensuring AI transparency. Despite the rhetoric against overregulation, federal agencies are obligated to assess and publicly register their AI systems. Failure to comply with these guidelines prompts agencies to cease their AI use.
Unearthing Meaningful Insights
As we prepare for the next phase of our research, we will critically analyze the content found within the AI transparency statements. Are they substantive? Do they adequately disclose risks and safeguards? Initial observations suggest significant variability in quality, raising pertinent questions about the real intent behind these statements.
If governments genuinely aspire to incorporate responsible AI practices, they must enforce their own policies rigorously. When even diligent researchers can’t easily find these transparency statements, it’s a clear indication that true transparency is still an elusive goal.
The journey towards responsible AI use in government is still unfolding, and the need for consistent and meaningful transparency remains more crucial than ever.
Inspired by: Source
