The potential abuse of consumer AI tools is raising concerns, with budding cybercriminals apparently able to trick coding chatbots into giving them a leg-up in producing malicious programs.