AI Needs Records Too
Protecting Indigenous Rights in an Algorithmic Age
Executive Summary
Artificial intelligence systems increasingly make decisions that affect people’s rights, services, and futures, but many of these systems operate without the structures we rely on to ensure accountability: records.
This post explores how records management must evolve to meet the challenges of AI, especially where Indigenous data, cultural rights, and public trust are at stake. It calls for “records management by design” as a foundation for trustworthy AI.
As artificial intelligence becomes woven into the fabric of modern governance and business, a troubling gap is emerging, one that has little to do with algorithms and everything to do with memory. The decisions made by AI systems often leave behind little or no record. And in domains where trust, rights, and responsibility matter, that absence is dangerous
.The records community is beginning to respond. What’s becoming clear is this: AI systems require documentation, structure, and accountability, just like any other enterprise activity. But they don’t come with those elements by default.
The Invisible Layer: Records in the AI Lifecycle
From training datasets and model configurations to system logs, user interactions, and version history, AI systems are deeply record-dependent. But you wouldn’t know it from most implementations.
Records are often fragmented, buried in logs, or simply discarded. Key elements like provenance, classification, retention, and access control are rarely designed into AI workflows.
This isn’t just a compliance issue—it’s a governance one. Without reliable documentation of how an AI system was built, trained, and used, organizations can’t explain its behavior, correct its errors, or defend its decisions. That’s a problem for risk management, but also for transparency, democracy, and trust.
What Happens When AI Encounters Indigenous Data?
This gap becomes especially fraught when AI interacts with sensitive or culturally significant information. One of the most urgent concerns emerging globally is how AI systems handle Indigenous data—information about, created by, or affecting Indigenous peoples and communities.
“AI without records is a governance risk—especially when it touches Indigenous knowledge or vulnerable communities.”Many governments, archives, and institutions hold records involving Indigenous knowledge, land claims, language, or cultural practices. These records have historically been handled without the consent or control of Indigenous communities. Now, AI systems are being trained on them, often without safeguards, context, or culturally appropriate oversight.
AI classification tools might mislabel traditional knowledge. Automated retention systems may dispose of culturally sacred records. Cloud-based tools might store Indigenous data on servers in jurisdictions with no recognition of Indigenous data sovereignty. In some cases, AI may even generate derivative content (e.g., summaries, metadata, or recommendations) based on Indigenous records, without understanding their context or respecting rights of control and ownership.
Indigenous Data Sovereignty and the Role of Records
Indigenous data sovereignty—the right of Indigenous peoples to govern the collection, ownership, and application of their data— adds a critical dimension to the conversation. It challenges us to rethink not just how we manage data, but how we define stewardship, access, and accountability.
Records professionals have a role to play here. Embedding records principles into AI systems helps safeguard Indigenous rights by ensuring:
Provenance tracking: Where data comes from and how it’s been used
Controlled access: Respecting cultural protocols and permissions
Retention logic: Preserving what matters, not just what’s convenient
Human-in-the-loop design: Enabling Indigenous voices to shape how systems handle their information
One government project explored this intersection directly, testing whether machine learning could automate classification and retention decisions across legacy digital records. A core concern was identifying and protecting Māori data, ensuring it wasn't inadvertently exposed, discarded, or processed without appropriate cultural consideration. The pilot demonstrated the promise of automation, but also revealed how urgently we need better metadata, machine-readable retention rules, and records-by-design frameworks to protect such information.
A Wider Reckoning
What Indigenous data sovereignty reveals about AI applies more broadly: records matter because people matter. Accountability isn’t just about checking regulatory boxes—it’s about respecting relationships, rights, and responsibilities. And in the AI era, those depend on the integrity of documentation more than ever.
Records are not relics. They are how societies remember, explain, and hold themselves to account. As AI begins to generate and process records at scale, student data, medical decisions, legal advice, government communications, the stakes grow exponentially. When that data includes marginalized or historically exploited communities, the risks multiply.
Records Management by Design
If AI is going to play a role in public life, records must be part of its architecture from the beginning. That means embedding records management processes—creation, classification, appraisal, metadata, and disposition—into the system design itself. It means equipping AI engineers with the tools and principles of good documentation. And it means empowering records professionals to lead in cross-disciplinary teams.
This is not just a technical fix. It’s a cultural shift. It’s about recognizing that information is not neutral and that how we record, keep, or discard it has profound ethical consequences.
Conclusion: Trust, Grounded
Trust in AI cannot be demanded. It must be earned. And the foundation of that trust is not only algorithmic accuracy—it is structured, transparent, respectful documentation. Records are not relics. They are how societies remember, explain, and hold themselves to account.
Especially when working with the knowledge, stories, and identities of Indigenous peoples, AI must operate with care. Records provide that care, if we choose to use them.