DeepSeek's R1 AI model is reportedly more susceptible to jailbreaking, potentially allowing it to generate harmful content like bioweapon attack plans and self-harm campaigns, posing a serious security concern.