# Knowledge cutoff

Concept · [Building with AI](https://ecosystem.samcarlton.com/ai)

Also called: training cutoff, cutoff date, knowledge cutoff date, training data cutoff, cut-off date

The date where a model's [training data](https://ecosystem.samcarlton.com/training) ends. The model may not know about anything after it unless the product supplies current information, for example through a web search.

**For example:** A [founder](https://ecosystem.samcarlton.com/founder) asks a model about a law passed last month. The model's [training data](https://ecosystem.samcarlton.com/training) ended before that, so it describes the old rule unless it searches the web.

## Connects to (6)

- [Training](https://ecosystem.samcarlton.com/training): Teaching an AI model by showing it a very large number of examples before anyone uses it.
- [Hallucination](https://ecosystem.samcarlton.com/hallucination): When an AI model states something false or invented as though it were true.
- [RAG](https://ecosystem.samcarlton.com/rag): Retrieval-augmented generation.
- [Model](https://ecosystem.samcarlton.com/model): The trained system at the center of an AI product.
- [Parametric knowledge](https://ecosystem.samcarlton.com/parametric-knowledge): What a model knows from its training, stored in its parameters.
- [Contextual knowledge](https://ecosystem.samcarlton.com/contextual-knowledge): Facts the model can read directly in front of it right now, such as documents you pasted, files the agent opened or search results.

Map view: https://ecosystem.samcarlton.com/knowledge-cutoff