What is a large language model?

Definition

A large language model is the engine inside tools like ChatGPT, Claude and Gemini. It's trained on a very large amount of text and learns to predict which words come next, which turns out to be enough to summarize, translate, draft and answer questions.

FINRA's Regulatory Notice 24-09 describes LLMs as a type of generative AI that uses deep learning and large data sets of language to identify, summarize, predict and generate text.

Why it matters for advisors

Most AI products advisors buy are built on someone else's LLM. A notetaker or CRM assistant may send your data to a model provider you never signed a contract with. That makes the model provider a likely sub-processor, which matters for vendor due diligence and Regulation S-P.

LLMs also have limits worth knowing: they don't look things up unless they're connected to a source, they can be out of date, and they can make things up. Knowing that helps you decide which tasks to give them.

Example

During due diligence on a notetaker, the CCO asks which LLM provider the vendor uses, whether client transcripts are sent to it, and whether that provider can keep or train on the data.

Watch out for

The same model can be safe or unsafe depending on the plan and contract. A consumer account and an enterprise plan of the same assistant can have very different data terms.