Grounding vs. training

Grounding means your data is retrieved and placed in front of a model at the moment of a request — it stays your data, in your store, and the model forgets it after answering. Training means your data is absorbed into a model's weights, where fragments can persist and resurface. Most practical questions about AI and company data resolve to which of the two is happening.

Why it matters

For a dealer evaluating AI-assisted software, the question "will our prices train someone else's model?" has a checkable answer: a grounded system with no-training terms in the provider contract does not feed your data into weights. In 2026 off-the-shelf models are good enough that client-data training is rarely needed at all.

Common confusion

"Not used for training" is not the same as "zero retention." They are different controls — one governs weights, the other governs storage — and a careful vendor names which ones apply, in the contract rather than the pitch.

Where we use it

The full mechanism, including provider data-control comparisons, is Where an AI quoting system gets its prices; our own commitments are in the AI Policy.