CORANZ Commentary
Government departments are increasingly embracing artificial intelligence (AI), promising greater efficiency, faster processing and improved public services. There are undoubtedly benefits, but an important question remains largely unanswered. At what point does AI cease being an administrative tool and begin influencing the policies and decisions for which elected representatives and public officials are accountable?
There are already signs of change in official correspondence. Responses from central and local government can be extraordinarily verbose, technically elaborate and filled with qualifications that almost overwhelm the original question. AI detection services such as GPTZero frequently identify characteristics associated with machine-generated writing, although their assessments are not conclusive proof. The answers may be there, but finding them can require considerable perseverance.
Using AI to improve grammar, organise information or prepare a draft response is one thing. Asking it to analyse evidence, evaluate public submissions, reach conclusions or recommend policy is quite another. AI can produce remarkably convincing arguments, but those arguments depend on the information supplied, the assumptions embedded in the system and the questions asked of it. An official seeking justification for an existing policy may receive a very different analysis from one genuinely exploring alternatives.
This is particularly relevant to environmental management and public access, where government agencies regularly receive submissions challenging established policies. AI could help officials examine large volumes of material, but it could also obscure important minority arguments or reinforce existing departmental thinking. A submission questioning the assumptions behind a proposal deserves genuine consideration, not merely an automated explanation of why those assumptions should prevail. The responsibility for exercising independent judgement must remain with identifiable human officials.
Interestingly, the Government already recognises these concerns. Its Public Service AI Framework includes principles of transparency, explainability and human accountability, stating that people receiving AI-assisted services should understand how AI is being used. The accompanying Responsible AI Guidance strongly recommends publishing information about agency AI use and maintaining an AI register. Yet the framework is explicitly non-binding, leaving an important distinction between recommended transparency and an enforceable disclosure requirement.
On 6 October 2026, Chief Ombudsman John Allen issued further guidance, acknowledging that AI can assist agencies with official information requests while emphasising that ministers, government agencies and local authorities remain accountable for their decisions. Another Ombudsman case concerned a request for a ministerial officeholder’s AI chat logs, including material potentially relating to government policy. These developments demonstrate that AI transparency is already a practical accountability issue, not simply a hypothetical concern about future technology.
CORANZ believes that whenever AI materially contributes to official correspondence, policy analysis, recommendations or decision-making reports, that involvement should be declared. The disclosure should explain what AI was used for, whether its conclusions were independently verified and which official accepts responsibility. Routine spelling or formatting assistance need not attract the same requirements, but substantive analysis and advice should never be concealed behind an official signature. The same principle should apply to local authorities, whose decisions can have equally significant consequences for communities and public resources.
Artificial intelligence has considerable potential to improve public administration, and its responsible use should be encouraged. But efficiency must not come at the expense of transparency, meaningful consultation or accountable human judgement. Government already expects the public to trust its decisions; it should be equally willing to explain how those decisions were reached. If AI is helping formulate government policy, the public has a right to know.
Disclosure. AI assisted in the research and development of this article