Characterising Deception in AI: A Survey