The study has thrown up hilarious, and concerning, findings. Some AI models cheated, others hacked into leaderboards. What should we make of it all?