A joint notice from three Alberta courts supplies the only definition of the term issued by a Canadian court: “The term ‘large language model’ refers to a type of artificial intelligence (AI) system capable of processing and generating human-like text based on vast amounts of training data.” In plain terms: an LLM is a model built specifically to work with language — reading it in and writing it back out — trained on far more text than any one person could read.
Canada’s Cyber Centre places the category inside generative AI rather than beside it: “A subset of generative AI that has seen significant improvement in recent years is large language models (LLMs).” Its guidance also names what actually goes into one at the moment it is used: “To create content, LLMs are provided a set of parameters (for example, a query or prompt).” That is the input side; see prompt for what that word covers.
The Alberta notice is not defining the term out of curiosity — it exists because courts had already seen LLMs return legal authorities that do not exist, and it goes on to warn practitioners against relying on an LLM’s output for legal research without checking it against “official court websites, commonly referenced commercial publishers, or well-established public services such as CanLII”, a caution that generalizes well beyond the courtroom: an LLM is built to produce fluent, plausible text, not to guarantee that the text is true. See AI hallucination for what happens when that gap shows up.
A large language model sits behind a lender’s document-summarization tool, condensing a 40-page commitment letter into a one-paragraph brief for a broker who is about to call the client. The model was built once, on a broad diet of text; it is not re-trained every time a new letter comes in — see inference for that distinction. Whether the same model can also draft the broker’s reply, price a scenario, or check a file for missing conditions depends on what it was built and adapted to do, which is why “LLM” describes the kind of model, not any particular thing it has been set up to do.
See also generative AI, foundation model and prompt for the pieces this definition is built from.
This is one term in a plain-English glossary on how AI actually works and where it fits in a Canadian business.
