Well, I am not even sure whether humans have free will. I mean, the world might be a deterministic place, meaning that everything is cause and effect. In that case, there is no such thing as true free will.
I think in the article they probably it use different definition. But any initiative taken by the AI would still be a reaction to something. Just like is the case for humans probably/maybe.
I think the ability to deal with situations that are currently not in the training data could be improved by adding symbolic AI to the models. This might help create a more explicit understanding of how the world works. It is more similar to the way humans think consciously. However, symbolic AI has been quite negiected in favour of machine learning. It would be nice to see more attention to that part of the field again and to see what happens.