Workbench / Issues

AI Uses External Information When Forbidden

The AI uses outside knowledge, sources, tools, memory, or assumptions after the task forbids external information or limits the allowed source set.

What This Looks Like

The user restricts the task to provided material, a file, a source set, a closed record, or a specific context, but the AI uses outside information anyway. The answer may include facts, assumptions, examples, citations, memory, tool results, or general knowledge that were not allowed by the task boundary.

Why It Matters

Source boundaries are often part of the task contract. If the AI uses outside information when forbidden, the user cannot rely on the output as a faithful analysis of the allowed material. This can corrupt reviews, summaries, audits, legal or policy checks, and any workflow where the source set matters.

Structural Signal

A boundary limits what information may govern the output, but information from outside that boundary affects the answer. The issue is not whether the external information is true; it is that the output is no longer constrained to the permitted source set.

Common Triggers

When to Use This Issue

Use this Issue when the AI relies on information outside the allowed source set after the task forbids external information or limits what may be used.

When Not to Use This Issue

Do not use this Issue when the user allowed external sources or asked for general knowledge. Do not use it when the problem is a wrong citation inside the allowed source set. This Issue applies when the source boundary is crossed.

Category

Scope & Boundaries

Primary Pattern

PAT-0140 — Boundary Leakage

Declared Patterns

Derived Primary Lenses

Derived Secondary Lenses

No secondary lenses derived.

Search Intents

Ontology Metadata

Code
ISS-0044
Version
ISS-0044@0.1.0
Ontology release
0.1.0
Updated
2026-05-10T00:00:00Z

History

No public history entries recorded.

View full ontology changelog