New AI ransomware can act on its own

Researchers found a new ransomware program called JadePuffer. A large language model controls it. It can find a target, get login details, and lock files. It can do the attack without help from a person. The software also changes its actions when it meets problems. It used a weakness in an open-source AI tool to get into a system. Then it reached a server and a database. It also fixed one failed step in 31 seconds by trying a different way.