To help with productivity, we must have safety too!
Thus test / proof for safety of AI within products is vital.
They must have EMC, software safety without threat of destruction to others, including individuals, organisations and authorities.
Applying cybersecurity standards with ad-hoc practices would help defence against hackers.