Presumably they would accept AI as a full epistemic participant if it was proven to be actually thinking. Which it isn't as of now.

It's a useful tool in the hands of experts, with serious threats to development of new expertise.