Correct, it looks like they tweaked the model's response to cover for Hoffman. There are many news reports about this, the answers are also very strangely opinionated. You can try the same questions in OpenAI to see something more accurate that's based on the data it was trained on and not later editorialization.