Analysis
What Is Generative AI Data Exfiltration? Definition
Publication date unknown · Discovered nhimg.org
Page published
Publication date unknown · First observed: 6 Nov 2026
Coverage timeline
Single-source analysis — one report is available.
Why it matters
Generative AI data exfiltration can occur through ordinary-looking responses and tool calls, making connector permissions and retrieval boundaries important defensive controls.
NHIMG’s glossary entry explains how generative AI systems can expose sensitive information through prompts, outputs, retrieved context, integrations, and agent tool calls. The entry outlines general exposure paths and mitigations, including scoped access, separated corpora, controlled connectors, and reduced prompt-visible context; it reports no specific incident or new experimental finding.