Guidance and examples to help you manage information and records created by AI systems, in line with public recordkeeping requirements.
Public information and AI
As you explore the use of artificial intelligence (AI) and emerging technologies, consider how these tools interact with existing information and records management obligations.
The use of AI does not change what constitutes a public record. The requirement to assess records based on their function, purpose and value does not change. Records must continue to be managed in line with the Public Records Act 2005 — regardless of the technology used to create or interact with them.
Refer to Archives New Zealand for authoritative guidance on retention, disposal and the long-term management of public records. The guidance helps public sector organisations to identify risks and considerations for information and records created or affected by using AI.
The following examples and scenarios show how records generated or used by AI may arise in practice.
These examples illustrate common situations to help encourage good information management behaviours. This does not replace formal recordkeeping advice.
For detailed advice on retention schedules, disposal authorities or agency-specific practices — consult directly with Archives New Zealand.
Scenario 1: AI summarisation tool used for internal meeting notes
A team uses an AI tool to summarise internal meetings based on audio recordings. The summaries are sent out for review and stored for future reference.
Records that may need to be retained:
the summary document, if it’s relied upon to inform decisions, follow-up actions or future work
any accepted manual edits or annotations made by staff
a brief record of the AI tool used, especially if concerns about accuracy or context arise.
If the summary replaces traditional meeting minutes or is used to support business decisions, it must be managed accordingly.
Scenario 2: Using a chatbot to support public enquiries
A public-facing virtual assistant or chatbot is deployed to answer general questions from users on a government website. It uses curated content and is monitored by staff.
Records that may need to be retained:
A representative sample of public interactions, especially where the chatbot replaces or supplements what a person would do.
The content source material used to train or support responses.
Documentation of the oversight processes, including regular content reviews and escalation procedures.
Any actual advice, guidance or information provided that meets the threshold for official communication, particularly where it influences public understanding or action.
Where AI responses are effectively delivering a public service, the interactions and supporting governance should be retained for transparency and accountability.
Advice or guidance provided through a chatbot may be considered part of the public record if it mirrors official service delivery.
Scenario 3: AI tools used in pilot or experimental settings
An agency runs a pilot project using an AI tool to explore whether it can reduce manual workload. The tool is not adopted, but insights are used for future planning.
Records that may need to be retained:
A record of the pilot’s purpose, scope and parameters.
Evaluation reports, feedback from users and summary findings.
A sample of outputs to show performance or limitations.
Any decisions or lessons learned that inform future initiatives.
Even if the AI system is not adopted, the pilot may generate valuable records about how new technology is being explored and evaluated.
Scenario 4: Use of generative AI (GenAI) to draft documents or advice
An advisor uses a GenAI tool to create a draft document. The advisor reviews and edits the content before finalising it.
Records that may need to be retained:
The final version of the document, as part of the normal recordkeeping process.
The AI-generated draft if it influenced or contributed meaningfully to the final version.
The prompt or instructions used to generate the draft, if necessary to understand how the content was produced.
Documentation of staff review and approval prior to submission.
This ensures transparency in the drafting process and maintains an auditable trail of how official advice was developed.
