The principle "Key requirements for trustworthy AI" has mentioned the topic "controllability" in the following places: