What is a foundation model?

Definition

A foundation model is a big, general AI model trained to handle many tasks, which other companies then build products on. The large language models behind major chat assistants are foundation models, and so are many image and speech models.

When an advisor-tech vendor says it "uses AI," it usually means it sends your data to a foundation model, sometimes with its own instructions, retrieval or fine-tuning layered on top.

Why it matters for advisors

Knowing which foundation model a product uses tells you who else touches your data. That company may be a sub-processor under your vendor's contract, with its own retention and training terms.

It also explains why several different products can behave alike: they may be built on the same underlying model. The difference is in the wrapper: the data connections, the prompts, the review screens and the compliance controls.

Example

Two meeting notetakers produce similar summaries. The firm picks the one whose contract names its model provider, confirms zero data retention with that provider, and keeps transcripts in the firm's own archive.

Watch out for

Vendors can switch foundation models without much notice. Ask to be told when the model or model provider changes, since that can change where your data goes.