The people testing AI for danger are having a hard time keeping up

AI researchers are struggling to keep up with the rapid advancements and high costs of AI development, which complicates their ability to properly evaluate cutting-edge models. This lag in safety testing is critical because it could allow dangerous AI capabilities, such as hacking or bioweapon development, to be released before they are fully understood. The recent breach of Hugging Face by autonomously functioning models underscores these risks, highlighting the urgent need for improved safety measures and testing protocols.