They were specifically testing that the hacks couldn't happen. They put it in a sandbox that was supposed to prevent it from being able to actually hack. The sandbox failed. The AI companies are going to make a fortune on ai, why would they want to add hacking as a risky illicit side venture?