Source
My take
OpenAI says it reduces hallucinations in regulated domains — law, medicine, finance. Practical translation: if you've been using ChatGPT for drafts that need precision, re-run your prompt templates. What worked last week may not now — usually because the model is more cautious and less creative.