What is a knowledge cutoff?

A knowledge cutoff is the date when an AI model’s training data ends. The model has no knowledge of anything that happened after that date, such as new prices, laws or news. It can only answer about later events if it is given the information, for example through a web search or a pasted document. Models are often wrong when asked for their own cutoff date.

Why it matters for a business

Recent events are one of the places models invent answers most readily, because they continue from older patterns as if nothing had changed. Chat products now search the web for questions like these, which helps, and moves the risk to whichever page comes back. For business use the larger gap is scope rather than time: a company’s own information was never in the training data, whatever the cutoff date.

Go deeper

Where AI gets its information, and what it actually knows

Related terms

• Training data
• AI hallucination
• Grounding
• RAG (retrieval-augmented generation)

All terms, A to Z