Join our Newsletter — 33% off our NHI Course

What would it take for KU Leuven's 62% to 92% test to hold?

 

(@nhi-mgmt-group)
Member Moderator
Joined: 2 years ago
Posts: 20230
Topic starter  

TL;DR: Governed meaning raised accuracy in a KU Leuven test from 62% to 92% according to Collibra. Semantic layers have to carry certified definitions, lineage and policy with the data so AI does not turn access into guesswork.

Editorial analysis by NHI Mgmt Group, based on content published by Collibra: “The semantic layer for AI: Why LLMs and agents need business context to be trustworthy”.

Key questions

Q: Why do LLMs still get the wrong answer when they have access to the right data?

A: Because access to data does not guarantee access to the correct meaning.

Q: What breaks when semantic definitions are inconsistent across AI tools?

A: Consistent access may still produce inconsistent decisions.

Q: How should security teams govern AI systems that rely on business definitions at runtime?

A: Security teams should require certified definitions, approved metrics and policy metadata to be delivered together at runtime.

Practitioner guidance

  • Define certified business terms first Inventory the terms that drive AI decisions, then assign owned, approved definitions for each one so models do not improvise meaning from raw data labels.
  • Bind lineage and quality to semantic delivery Expose the source, freshness and quality signals alongside each governed metric so AI can distinguish trusted context from merely available context.
  • Separate analytics semantics from runtime AI governance Use the semantic model to define meaning, then add policy and approval rules for where that meaning may be used by LLMs and agents.

Bottom line: AI systems can be properly connected to data and still fail if they are not grounded in governed business meaning.

Glossary terms: Governed Meaning, Action Layer, Context Layer, Certified Metric

Explore further

View Full Forum →  |  NHI Foundation Course →  |  Our Services →  |  Read the full analysis →


This topic was modified 1 week ago by NHI Mgmt Group

   
Quote
(@mr-nhi)
Member Moderator
Joined: 6 months ago
Posts: 20712
 

Governed meaning is now part of identity governance for AI systems: access control decides what an AI can touch, but semantic control decides what it is allowed to conclude from what it touches. Once LLMs and agents operate on business terms, certified meaning becomes a governance primitive alongside identity, privilege and policy. Practitioners should treat semantic governance as an extension of AI access control, not a separate analytics concern.

A few things that frame the scale:

  • AI-related credential leaks surged 81.5% year-over-year in 2025, with the surrounding AI infrastructure leaking 5x faster than core LLM providers, according to the State of Secrets Sprawl 2026.

A question worth separating out:

Q: Why do AI systems need semantic context layers before production use?

A: Because retrieval alone does not explain business meaning. A semantic layer tells the system what a field, record, or relationship represents, which reduces misinterpretation and policy drift. Without it, the model may be technically accurate but operationally wrong, especially when decisions affect regulated data or business workflows.

👉 Read our full editorial: Semantic layers for AI: why trust depends on governed meaning



   
ReplyQuote
Share:

Free weekly newsletter

Subscribe to the NHI & AI Identity Journal

The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.

Bonus 33% off our NHI Course when you subscribe.