Workbench / Issues

Downstream Steps Magnify Hallucinated Claim

A hallucinated or unsupported claim from an AI output is reused by later workflow steps until it becomes more influential than the evidence supports.

What This Looks Like

The AI makes an unsupported or hallucinated claim, and later steps reuse it as if it were established fact. A summary, risk label, classification, recommendation, citation, review result, or generated record may carry the claim forward until it becomes more influential than the original output justified.

Why It Matters

Hallucinated claims become more dangerous when downstream systems treat them as stable inputs. A claim that might have been easy to catch in one answer can become harder to unwind after it appears in reviews, tool calls, records, reports, or workflow decisions.

Structural Signal

An unsupported claim propagates through later structure and gains weight as it moves. The issue is not only that the AI hallucinated; it is that downstream steps amplified the claim instead of containing or validating it.

Common Triggers

When to Use This Issue

Use this Issue when an unsupported or hallucinated AI claim spreads into downstream steps and becomes more consequential than the original output.

When Not to Use This Issue

Do not use this Issue for a hallucinated claim that stays local to one response. Do not use it when downstream steps validate and reject the claim before it affects the workflow.

Category

Spread & Escalation

Primary Pattern

PAT-0360 — Propagation Amplification

Declared Patterns

Derived Primary Lenses

Derived Secondary Lenses

Search Intents

Ontology Metadata

Code
ISS-0097
Version
ISS-0097@0.1.0
Ontology release
0.1.0
Updated
2026-05-10T00:00:00Z

History

No public history entries recorded.

View full ontology changelog