AI Disclosure in Public Procurement: What Buyers Are Asking and Why It Matters
The policy shift is already underway
In December 2025, the Digital Transformation Agency updated its Policy for Responsible Use of AI in Government, strengthening accountability measures, transparency requirements, and guidance for how AI should be governed across the Australian Public Service. Earlier in the year, the DTA published AI Model Clauses (version 2.0) to help government purchasers manage vendor relationships where AI is part of the tendered solution.
The APS AI Plan 2025, which sits above both of these instruments, sets out procurement transparency requirements and establishes a central register of generative AI assessments on GovAI, allowing agencies to share completed security, foreign ownership, and impact assessments and streamline procurement decisions.
Legal commentary is keeping pace. Sparke Helmore now recommends that agencies check whether AI is being used in tender responses and assess whether any unmanaged AI risk needs to be mitigated in upcoming procurements. For non-AI-based procurements, agencies are advised to check whether AI may form part of the tendered solution and whether market testing documents should include questions that identify AI use and assess the tenderer’s risk management approach.
In New Zealand, the Public Service AI Framework was released in January 2025, followed by updated Responsible AI Guidance for GenAI in February 2025, covering procurement among other areas. The approach is principles-based, emphasising transparency, accountability, and human oversight. New Zealand’s National AI Strategy, launched in July 2025, provides the broader strategic direction.
The trajectory across both markets is consistent. Disclosure is forming as an expectation. The pace and formality differ, but the direction is shared.
Why now?
The timing is not arbitrary. As explored in the first article in this series, AI has reduced the effort required to produce a tender submission. Volume has increased. Evaluation workloads have grown. And, as addressed in our second joint industry article, accountability for the claims made in AI-assisted responses remains with the submitting organisation regardless of how those claims were generated.
Disclosure sits between those two realities. It is the mechanism through which buyers can understand how a response was produced, and through which suppliers can demonstrate that their submission process is thorough regardless of the tools used to produce it.
The analogy is straightforward: understanding whether AI was used in a response is comparable to understanding whether an external bid writer was used. Both are legitimate. Both warrant visibility.
Buyers are using AI too
The disclosure conversation is often framed as something buyers require of suppliers. In practice, AI adoption is happening on both sides of the table.
Procurement teams are using AI to draft specifications, scope documents, and evaluation criteria. The DTA has conducted proof-of-concept trials using generative AI to assess technical case studies in its panel tender process, testing whether AI could assist human evaluators in future procurements. The Department of Finance has stated that evaluations will continue to be the responsibility of human evaluators, but that AI may be paired with evaluators to make future processes more consistent and efficient.
Chris Sullivan, CPO at Ramsay Health Care and former Group CPO at Coca-Cola Amatil, put it directly in The Bid Room: Perspectives Series: [AI] “amplifies good work or good relationships, and it can equally amplify disasters.”
The observation applies to both sides. If AI amplifies what is already there, then disclosure is how both buyers and suppliers understand what they are amplifying. An organisation with a thorough submission process will produce stronger work with AI assistance. One without it will simply produce unverified claims faster. Disclosure helps evaluators understand which they are looking at.
What disclosure achieves
The value of disclosure is practical, not performative.
For buyers, it provides a signal about how a response was assembled. It does not tell them whether the content is accurate. It does not replace evaluation. But it adds a layer of information to the assessment that helps procurement teams understand how a submission was assembled and what verification sits behind it.
For suppliers, disclosure is an opportunity to demonstrate the quality of their submission process. An organisation that can state clearly how AI was used, where human review was applied, and how claims were verified is communicating something meaningful about its internal processes. That signal matters in an environment where evidence of capability is increasingly important.
This distinction is likely to sharpen over time. As procurement teams become more experienced in evaluating AI-assisted submissions, the questions will move from general (“did you use AI?”) toward specific (“how did you verify the staffing model proposed in Section 4?”).
What this means for suppliers
Buyers asking this question means that others in your market are likely using AI to respond to similar tenders. The practical implications are straightforward.
As adoption of AI tools increases, suppliers operating in Australian and New Zealand public procurement should anticipate AI disclosure questions appearing with increasing frequency. The policy frameworks are in place. The legal guidance is pointing agencies toward asking these questions. It is a matter of when, not whether, disclosure becomes a standard feature of tender documentation.
The organisations that are best positioned are those that already have structured review processes around their tender submissions. A clear internal record of who drafted each section, who reviewed it, and what verification was applied to factual claims is standard practice in mature bid operations. AI does not change that requirement. It simply means there is one more production method to account for within an existing quality assurance workflow.
Equally relevant is the type of AI being used and how it is configured. AI that draws on verified response libraries, maps claims to actual delivery evidence and operates within the compliance structure of the submission is a fundamentally different tool from AI that generates content from general training data. The distinction matters to evaluators even if it is never explicitly asked about.
Suppliers with disciplined submission processes will find disclosure straightforward. Those without them were already exposed. AI just makes it more visible.
What this means for buyers
For procurement teams, the practical question is straightforward: does this submission hold up under scrutiny? The tool used to produce it is secondary to whether the claims are supported by evidence, verified by the supplier, and defensible under evaluation. That has always been the standard. In this context, AI just means a weak submission can be assembled faster than it used to be.
As Sullivan noted in the same session, the details of a submission reveal more than suppliers realise: “[These can be] very small cosmetic things, but it actually tells a massive story behind the scenes.” The same principle extends to AI-assisted content. How a response was produced is visible in its structure, consistency, and evidence density, whether or not the supplier intended it to be. Experienced evaluators are already reading these signals. Disclosure simply puts a name on what they are seeing.
The emerging norm is not about whether AI was used. It is about whether the submission, however it was produced, is trustworthy.
The emerging standard
As AI adoption increases, AI disclosure in procurement is moving from optional to expected. Australia is building the policy infrastructure. New Zealand has established the principles. The Australasian Procurement and Construction Council is working on how to extend these transparency expectations across government procurement more broadly.
For the tendering ecosystem, this is a process maturation. Disclosure normalises AI use while establishing transparency as the baseline expectation. It protects buyers by providing visibility into how submissions are produced. It protects suppliers by creating a structured way to demonstrate responsible practice.
The organisations, on both sides of the table, that engage with this shift early will set the standard for how AI-assisted procurement operates. Those that wait for disclosure to become mandatory will find themselves responding to requirements they had no hand in shaping.
