Knowledge Cutoff
The date after which an AI model knows nothing, because its training stopped there. Everything the model learned came from data gathered up to that point, so events, releases, and news from afterwards fall outside its knowledge. This point in time is called the knowledge cutoff.
A knowledge cutoff is the moment a model’s learning was frozen. An AI is trained on a large collection of text gathered up to a certain date, and after that the training stops. From the model’s point of view, the world pauses there. It can discuss anything up to the cutoff, but events after it are a blank.
This explains a very common surprise. Ask a model about something that happened last week and it may have no idea, or worse, it may hallucinate a plausible-sounding answer to fill the gap. It is not being evasive. That information was never part of what it learned.
There is an important twist to know about. Some AI tools can now look things up on the web or in your documents while they answer, which lets them work with information newer than their cutoff. This is often done through a technique called retrieval-augmented generation. When a tool can search, its answers can be current; when it cannot, it is limited to what it learned during training.
The practical habit is simple. For anything recent, such as prices, news, current office-holders, or the latest version of a product, check whether your tool can search the web, and treat unsourced claims about recent events with extra care.