Comment by dzaima

Humans can hallucinate up some API they want to call in the same way that LLMs can, but you don't call all human mistakes hallucinations; classifying everything LLMs do wrong as hallucinations would seem rather pointless to me.