IM BLOG: NAA Check-up: AI Is Everywhere - So Why Is It Missing from Records Management?

A wake-up call for information management

The National Archives of Australia’s Check-up 2025 Whole-of-Government Summary Report provides one of the clearest pictures yet of how Australian Government agencies are managing records, information and data. Check-up is the NAA’s annual information management self-assessment survey, aligned to the Information Management Standard for Australian Government and the Building trust in the public record policy. The 2025 survey was conducted between September and December 2025 and covered 175 agencies.

The headline result is positive: overall information management maturity reached 3.72 out of 5, continuing a steady improvement from 3.60 in 2022, 3.65 in 2023 and 3.70 in 2024. The strongest areas remain creating information assets at 4.37 and storing, preserving and managing information assets at 4.13. But the weaker areas remain familiar: use, reuse and interoperability at 3.37 and appraising and disposing at 3.28.

For RIMPA Global members, the most important insight may not be the overall maturity score. It is the gap between the widespread use of AI across government and the very limited use of AI in records and information management.

The AI adoption gap in records management

 The Check-up report found that 87% of agencies reported using AI, while only 15% used it for information management and 8% used it to create metadata. Sixty-two per cent consulted information management teams when integrating AI into operations.

However, the 87% figure measures adoption only. The survey does not establish whether agencies understand which AI-generated inputs, outputs or decisions constitute Commonwealth records, or whether they are being captured and managed appropriately. This is an important consideration, not a Check-up survey finding.

The distinction matters because agencies continue to face challenges with metadata, interoperability, sentencing, disposal, archival transfer, legacy systems and digital volumes. Carefully governed AI could assist in these areas, but only alongside Commonwealth records guidance, professional oversight and accountable information governance.

The metadata problem is now an AI readiness problem

The Check-up report shows that only 24% of agencies have a fully implemented metadata strategy, although a further 43% have one under development. It also found that only 32% of agencies usually or always undertake quality assurance checks on metadata consistency, and only 31% usually or always ensure data exchanges with other agencies are accompanied by metadata or data dictionaries.

These are not just records management issues. They are AI readiness issues. AI systems depend on context. If records and data are poorly described, inconsistently tagged, stored across uncontrolled locations, or separated from the metadata that explains their origin and status, AI tools may retrieve the wrong information, summarise incomplete evidence, or provide outputs that appear authoritative but cannot be trusted.

In that sense, metadata is no longer a back-office concern. It is part of the trust layer for digital government.

Digital volume is growing faster than governance capacity

The report also highlights the scale of the information challenge. Agencies estimated they held more than 1,028,246 terabytes of digital information assets in 2025, more than double the 424,902 terabytes reported in 2024. Physical information asset volumes fell to their lowest level in six years, but digital volume is growing rapidly.

At the same time, the report shows ongoing challenges in sentencing and disposal. Only 49% of agencies sentenced digital information assets in the relevant reporting period, meaning 51% had not sentenced any digital assets. Agencies also identified the volume of digital information, lack of resources, lack of available staff with necessary skills, and difficulty sentencing digital assets as persistent challenges.

This is where AI could be transformative - not as a replacement for professional judgement, but as an accelerator for classification, clustering, duplicate identification, metadata suggestion, defensible disposal workflows and risk identification. Used responsibly, AI could help agencies understand what they hold and reduce the burden on already stretched records teams.

Appraisal and disposal remain the hard work

 Appraising and disposing is the lowest-performing maturity area in the NAA Check-up report, scoring 3.28. However, its three headline statistics require context before being treated as evidence of poor information management.

Only 36% of agencies facilitate automated identification of assets due for destruction or transfer. It is unclear whether this covers digital holdings suitable for automation or all holdings, including physical and legacy records. If it includes all holdings, automation across eligible digital information may be higher.

Similarly, only 34% promptly destroy temporary assets after their authorised retention period. Delays may reflect disposal freezes, legal holds, continuing business needs, pending approvals, budget constraints or competing priorities, rather than incapacity.

The 17% rate for transferring national archives within 15 years may also reflect retrieval costs, preparation effort, ongoing access needs and reluctance to relinquish custody.

AI can help identify candidates, group similar assets and support professional review. However, it cannot resolve legal, operational, funding or custody decisions. The key question is whether delayed disposal or transfer is conscious, authorised and regularly reviewed.

AI governance must include records governance

The National Archives has already issued guidance on information management for records created using AI technologies, including how to identify and manage Commonwealth records created by, or relating to, AI systems. The guidance recognises that AI-generated outputs, prompts, inputs, system documentation, implementation records and governance materials may all form part of the recordkeeping environment.

That guidance is highly relevant to the Check-up findings. If 87% of agencies are already using AI, then agencies need to know what records are being created by those systems, what evidence must be retained, and how AI-supported decisions can be explained later.

For RIMPA Global members, this is a critical point. AI governance cannot be separated from information governance. Every AI project should ask: What information is the system using? Is it authoritative? What records are created? What metadata is required? Who can access the information? How long should it be retained? Can the decision be reviewed?

The National AI Plan raises the stakes

Australia’s National AI Plan, released in December 2025, sets out three goals: capturing the opportunity of AI, spreading the benefits through adoption and improved services, and keeping Australians safe through regulatory and legislative frameworks that mitigate harms.

The establishment of the Office of AI in the Department of the Prime Minister and Cabinet on 15 July 2026 further signals that AI is now a whole-of-government priority. The Office of AI is tasked with coordinating across Australian Government agencies to design and legislate a new Australian AI standard, including requirements for large AI data centres and copyright protections for Australian creators.

This national momentum raises the stakes for records and information management. Agencies will need more than AI pilots and strategy documents. They will need evidence that their information is governed, accurate, accessible, secure, appropriately retained and capable of supporting reviewable decisions.

Legislation is moving toward transparency and accountability

The broader legislative environment is also changing. In Australia, the Privacy and Other Legislation Amendment Act 2024 introduced reforms including increased transparency about automated decisions that use personal information. These automated decision-making transparency obligations are scheduled to commence on 10 December 2026.

Australia’s Digital ID Act 2024 also strengthens the governance environment for digital services by establishing accreditation, privacy and oversight arrangements for the Australian Government Digital ID System. While it is not an AI law, it reinforces the same principle: trusted digital government depends on clear governance, strong controls and accountable information practices.

In New Zealand, the Privacy Amendment Act 2025 introduced Information Privacy Principle 3A, which came into force on 1 May 2026 and requires agencies collecting personal information indirectly to take reasonable steps to notify individuals unless an exception applies. New Zealand has also introduced the Biometric Processing Privacy Code 2025, which sets specific privacy rules for biometric processing, including facial recognition and similar technologies.

A regional challenge for Australia and New Zealand

New Zealand’s AI Strategy: Investing with confidence focuses on accelerating AI adoption and innovation while supporting responsible use. Archives New Zealand has also made clear that outputs created using AI are public and local authority records, and that information recording the use of AI must be documented so records remain full, accurate, accessible and understandable in context.

Across Australia and New Zealand, the direction is consistent: AI adoption is accelerating, but accountability obligations are not disappearing. If anything, they are becoming more important.

What this means for RIMPA Global members

The Check-up report should be read as both a progress report and a call to action. Government agencies are improving their information management maturity, but the low use of AI in information management shows a significant opportunity gap.

RIMPA Global members have a central role to play. Records and information professionals should be at the table when AI strategies are developed, when systems are procured, when metadata models are designed, when AI use cases are approved, and when agencies assess risks around privacy, access, disposal and accountability.

The opportunity is not to automate records management blindly. The opportunity is to use AI in a governed, transparent and defensible way to strengthen the foundations of public administration.

From AI adoption to AI accountability

The NAA Check-up 2025 report shows that AI is already present across government. The challenge now is to make sure it is applied where it can improve trust, not just productivity. If AI can help agencies better classify information, improve metadata, identify records due for disposal, support transfer to archives, and improve discovery and reuse, then it can help address some of the most persistent information management challenges across government.

But AI will only succeed if it is grounded in good information governance.

For RIMPA Global and the profession, this is the moment to lead. AI may be the technology transforming work, but records and information management will determine whether that transformation is trustworthy, accountable and enduring.

Practical opportunities for agencies

Reference links

• NAA Check-up survey page: https://www.naa.gov.au/information-management/check-up-survey
• Check-up 2025 Whole-of-Government Summary Report: https://www.naa.gov.au/sites/default/files/2026-06/Check-up-2025-whole-of-government-summary-report.PDF
• National AI Plan: https://www.industry.gov.au/publications/national-ai-plan
• Office of AI | Department of the Prime Minister and Cabinet: https://www.pmc.gov.au/domestic-policy/office-ai
• Privacy and Other Legislation Amendment Act 2024: https://www.legislation.gov.au/C2024A00128/asmade
• Digital ID Act 2024: https://www.legislation.gov.au/C2024A00025/asmade
• NAA guidance: Information management for records created using AI technologies: https://www.naa.gov.au/information-management/manage-information-assets/types-information/information-management-records-created-using-artificial-intelligence-ai-technologies
• Archives New Zealand: Artificial intelligence and public records: https://www.archives.govt.nz/manage-information/how-to-manage-your-information/implementation/artificial-intelligence-and-public-records
• New Zealand Privacy Amendment Act 2025: https://www.legislation.govt.nz/act/public/2025/0053/29.0/whole.html
• New Zealand Biometric Processing Privacy Code 2025: https://www.privacy.org.nz/privacy-principles/codes-of-practice/biometric-processing-privacy-code/

Meet your blog author:

 

Brandon Voight

Brandon Voight is a Fellow at the Future Government Institute and Director of Enterprise Information Management - OpenText Australia & New Zealand. 

Originally a corporate lawyer, and now with over twenty years of experience in empowering customers through technology, his mission is to help OpenText customers unlock the value of their data, improve information governance, and drive informed decision-making.