British officials say AI chatbots could carry cyber risks
Academics and researchers have repeatedly found ways to subvert chatbots by feeding them rogue commands or fool them into circumventing their own built-in guardrails.
