Apple Intelligence protections bypassed with prompt injection - 9to5Mac
…strategies to get Apple’s on-device model to execute attacker-controlled instructions through prompt injection. Interestingly, they successfully executed the exploit without being 100% sure of how Apple’s local model…