The working group's research identified 5 different attack patterns applicable to any Artificial Intelligence system: