I agree with the major premise of the article: explainability of actions and recommendations by AIs will be important, but very difficult to active. Interesting about that law that takes effect in EU in a year!
There are ways to make AI decisions explainable. There just aren't that many people working on it. Darpa advertised a grant last year for this, but afaik it's the only group I've heard of tackling that problem space. As the field matures it will happen, no question.