Notifications
Clear all
Topic starter
23/06/2026 7:01 pm
TL;DR: AI systems fail less because they lack data than because they lack context, according to Collibra’s analysis of metadata management across structured and unstructured content. The governing problem is that AI can confidently retrieve the wrong answer when ownership, freshness, sensitivity, and policy context are missing, making metadata the control layer that turns content into defensible AI input.
NHIMG editorial — based on content published by Collibra: Metadata management is the missing layer that makes AI actually work
By the numbers:
- This world of unstructured data is where many AI projects start to slow down, and it is 80% of all enterprise data.
Questions worth separating out
Q: How should security teams govern AI use of unstructured content?
A: Security teams should require metadata that describes ownership, freshness, sensitivity, approval state, and intended use before unstructured content enters AI retrieval workflows.
Q: Why does metadata matter more when AI uses both structured and unstructured data?
A: Because AI does not respect the separation between databases and documents.
Q: How do teams know if AI content governance is actually working?
A: Look for traceability, policy consistency, and refreshed context.
Practitioner guidance
- Classify content by business use and sensitivity Map unstructured content into policy-relevant categories before AI systems can retrieve it.
- Apply active metadata to high-change content Prioritise documents, policies, tickets, and transcripts that change frequently or feed AI workflows directly.
- Separate search success from governance success Measure whether AI outputs are both relevant and authorised, not merely accurate-looking.
What's in the full article
Collibra's full blog post covers the operational detail this post intentionally leaves for the source:
- How Collibra frames metadata enrichment across documents, tickets, policies, and other unstructured content.
- The way active metadata is positioned to keep governance signals current as content, ownership, and use cases change.
- The distinction the vendor draws between traditional cataloging and metadata that actively supports AI workflows.
- The specific value proposition for Deasy Labs in automating discovery and classification of unstructured data.
👉 Read Collibra's analysis of why metadata management is the missing layer for AI →
Metadata management and AI context: what IAM teams should watch?
Explore further
25/06/2026 2:47 am
Metadata drift is becoming an AI governance failure mode, not just a data-management issue. AI projects slow down when content exists but the context does not travel with it. That is a governance problem because the system can retrieve material that is current in storage but obsolete in business meaning. The implication is that teams must stop treating metadata as cataloguing and start treating it as a control surface for authorised AI use.
A few things that frame the scale:
- The average estimated time to remediate a leaked secret is 27 days, despite 75% of organisations expressing strong confidence in their secrets management capabilities, according to The State of Secrets in AppSec.
- 43% of security professionals are concerned about AI systems learning and reproducing sensitive information patterns from codebases.
A question worth separating out:
Q: What is the difference between metadata management and simple content search?
A: Search finds content. Metadata management tells you whether that content is current, owned, sensitive, and approved for the intended use. In AI programmes, that difference matters because a retrievable answer is not necessarily a governable answer. Metadata is the layer that makes retrieval defensible.
👉 Read our full editorial: Metadata management is the missing layer behind trustworthy AI
