In plain English
A foundation model is a large AI model trained on broad, general data that can be adapted to a wide range of tasks through prompting or fine-tuning. Rather than building a new model for every job, developers start from a capable general model and specialise it. Large language models like GPT and Claude are the best-known examples.
Why it matters
Most AI products you use are thin layers over a handful of foundation models. That shared dependency shapes their strengths, limits, and risks.
A worked example
A legal-research tool, a coding assistant, and a customer-service bot might all be built on the same foundation model, each adapted with prompts and extra data.
Common confusion
A foundation model is not a finished product. It is the general engine; the product is what a developer builds and configures on top of it.