Supply Chain
Third-party components — models, libraries, datasets, plugins — can be compromised, backdoored, or vulnerable, and the weakness propagates to you.
What it means
An LLM application is assembled from many external parts: foundation models, SDKs, open-source libraries, and sometimes third-party datasets or plugins. Supply-chain risk is a weakness or tampering in any of those parts flowing through to your application.
Why it matters
The more external components an application pulls in, the larger its attack surface — and a single compromised dependency can affect everyone downstream of it.
How we approach it
We keep the external footprint deliberately small and the model-provider set short and named, so you can see exactly which providers requests reach.
Dependencies are vetted and kept current as part of release.
This is explicitly shared responsibility: you own the components inside your own application, and we own ours. For an embedded assistant, both sides have a supply chain to keep clean.
Frequently asked
Which model providers do requests reach?
A short, named set of established foundation-model providers — you can see which one is configured for each application, and requests go only there.
Who owns supply-chain security in an embedded product?
Both sides. The platform maintains the components it ships; the application embedding the assistant maintains its own dependencies and plugins.