Does AI (like Claude or ChatGPT) still commonly hallucinate or give confidently wrong answers?

One member noted that early on, especially with ChatGPT, it would confidently give wrong answers (e.g., wrong dollar thresholds) and you'd have to correct it before it agreed you were right. Another member said they use it constantly and haven't seen major hallucination issues recently, though they can't be sure it always reads everything correctly. The group generally felt this was less of a problem than it used to be, though not something to blindly trust.