Published 10 June 2026 · Syndie Bobeechun, Founder · Doreva Articles

Using AI on a government tender? Here is what the buyer can now ask you

The short answer: there is no general legal duty in Australia to disclose AI-assisted drafting in a tender response. But the agencies marking your bid now work under binding AI accountability rules with deadlines through 2026, some RFx documents already ask directly, and the strongest position is being able to answer six specific questions about how your response was prepared.

The RFT lands, and somewhere in the returnable schedules there is a question your team has not seen before. It asks whether artificial intelligence was used in preparing the response, and if so, how.

Five years ago that question did not exist. From here on, you should bid as if it is coming, because the rules on the buyer's side of the table have already changed.

What actually changed, with dates

Three things moved between late 2025 and now, and they are worth knowing precisely rather than vaguely.

The Commonwealth rewrote its own AI rulebook. The Digital Transformation Agency's Policy for the responsible use of AI in government, version 2.0, took effect on 15 December 2025. It requires agencies to designate accountable AI officials, keep a register of their AI use cases, and run impact assessments over them. Under the phased implementation timetable, the first of the new mandatory requirements starts on 15 June 2026, and the remainder, including mandatory AI impact assessments for in-scope use cases, are due by December 2026.

Procurement guidance arrived with it. The policy update shipped alongside an AI impact assessment tool and procurement guidance for agencies buying AI-enabled systems. The buying process itself is now governed.

Privacy law caught up. From 10 December 2026, amendments to the Privacy Act require organisations that use automated decision-making involving personal information to disclose it in their privacy policy. This one binds businesses directly, not just agencies.

Agency obligations, not bidder obligations. But the people marking your bid work under them.

None of these rules tells a bidder what to do with AI in a tender response. There is no general legal duty to disclose AI-assisted drafting. But that is not the end of the story, because rules like these change behaviour beyond their formal reach.

Why this reaches you, the bidder

The evaluator reading your submission now works inside an organisation with an accountable AI official, an AI use-case register, mandatory AI training and an impact assessment process. Their own use of AI is documented and reviewable.

People hold others to the standards they are held to. When an agency has to defend its own AI posture internally, its questions to suppliers get more precise. Some RFx documents already ask directly whether AI was used in preparing the response. Others reach the same place through originality, accuracy or confidentiality clauses. Many are silent today and will not stay silent.

The practical consequence: the bid teams that do well from here are not the ones that avoid AI, and not the ones that use it quietly and hope. They are the ones that can answer a sharp question in plain terms.

The six questions to be ready for

Read the policy, the assessment tool and the procurement guidance together and the supplier-side expectations converge on six questions. Treat them as a readiness check for your own bid process.

Did you disclose it? If the RFx asks about AI use, answer exactly what it asks. If it is silent, a short factual statement is often safer than silence. The bidder who is transparent should not lose to the bidder who says nothing, and evaluation panels increasingly know it.

Can you trace your claims? Every figure, project name and capability claim in the response should lead back to something real: a delivered project, a contract, a report. AI makes it faster to produce confident sentences. It does not change what an assessor will probe.

What stops fabrication? Generic tools will fill an evidence gap with something plausible. In government work a fabricated dollar value or an invented project name is not a quality problem, it is a probity problem. Whatever process you use needs a control that stops unsupported claims reaching the document.

Is there a record? If a panel, a probity adviser or a disappointed competitor asks how the response was prepared, a simple record answers it: what was used, which sections it touched, who reviewed them, who signed off.

Did a human own it? AI-assisted is fine. AI-submitted is not. Someone accountable needs to have reviewed every section before it went out, and the company that signs the response owns every word of it.

Where did the data go? Tender documents carry confidentiality obligations. Responses carry pricing, client information and personal details. Before any of that goes into a tool, you should be able to say where it is processed, whether the provider can retain or train on it, and what you keep out.

If your team can answer all six in one breath, AI is an advantage in your bid process. If it cannot, the gap is the process, not the technology.

How Doreva answers them

Doreva is a tender response platform built for Australian government work, and these six questions shaped how it was built.

Every exported document includes an AI Use Statement, automatically. The buyer is never surprised, and your team never scrambles to write a disclosure at 11pm before close.

Drafting is grounded in your own material: your past proposals, your case studies, your project history, your pricing inputs. Generation runs through a plausibility gate that refuses quantified claims with no source behind them. No invented project names. No made-up dollar values. No capability claims your evidence does not reach.

Every section keeps a version history: what was generated, what a person changed, who approved it and when. Imported past proposals are treated as confidential by default. And nothing exports without human review, because the submission belongs to the people who sign it.

Doreva does not replace your bid team. It gives them drafts they can defend, and the answers ready before the question is asked.

What to do this quarter

Three moves, none of which require a 40-page framework.

Write down your one-page AI rule set: approved tools, what never goes into them, who reviews, who signs. Decide your default disclosure wording now, before an RFx forces a rushed version. And keep a simple file note per bid recording what was used and who reviewed it.

If you are still weighing where AI fits in your bid process, start with our take on whether AI will replace tender writers. And if Commonwealth bidding is new to your team, the AusTender response guide covers the process end to end.

The bar for AI in tenders is rising on a published timetable. The bidders who treat that as a checklist item will be fine. The bidders who treat it as a way to stand out will win work with it.