CyberGuard dictionary / AI Security
Model Supply Chain
Model supply chain security concerns the origin and integrity of models, datasets, dependencies, tools, and build steps used in an AI system.
In simple terms
An AI application depends on more than its own source code. A model download or data feed can introduce a harmful change if it is not verified.
Technical explanation
Supply chain controls include source inventory, pinned versions, artifact integrity checks, trusted serialization formats, restricted build permissions, and review of updates. Risks may arise before deployment or during retrieval and tool use.
The chain includes model weights, datasets, code, packages, hosting, and updates. Each component can introduce untrusted behavior, so provenance and controlled promotion matter throughout deployment.
Examples in practice
A team upgrades to a model package from an unverified repository and its loader runs unexpected code during startup.
A new model package requests an unexpected dependency during installation. The team pauses release, checks the package source and build manifest, and uses a verified version.
Why it matters
Knowing where every AI component came from makes investigation and rollback possible when a dependency changes.
VERIFY & LEARN MORE
Sources and further reading
- genai.owasp.orghttps://genai.owasp.org/llm-top-10/↗
- genai.owasp.orghttps://genai.owasp.org/llmrisk/llm042025-data-and-model-poisoning/↗
Published 2026-10-15 · Last updated 2026-10-15 · CyberGuard editorial content. Source links provide context; they do not imply endorsement.