Ungrounded Content

Kind: anti-pattern

Layer: Enforcement Core

Record: architecture:ungrounded-content

Severity: discouraged

Scope: model_governance

Formed by: Generate answers, classifications, plans, or decisions without evidence retrieval, source references, confidence limits, or unsupported-claim handling.

Canonical: Ontology

A defect in which the model produces answers or decisions without retrieved evidence or disclosed uncertainty.

Listed in Architecture principles, after Prompt Sprawl and before Model Version Ambiguity.

Contracts

• Agent Creator Kernel

Refactored by

• Retrieval-Augmented Generation (RAG)
• Evidence Citation
• Claim Validation
• Abstain or Disclose Uncertainty

Severity

• discouraged

Category

• Anti-patterns

Force

• Model governance

Negated by

• Agentic Architecture

Linked from

• It looked right
• Agents as executed contracts
• The author is probabilistic
• Anti-patterns
• Model Architecture
• Model Architecture
• Agent creation
• Severity levels
• The forces