Researchers find AI-run ransomware

Researchers have found what they describe as the first agentic ransomware to be fully managed by a large language model. The software, JadePuffer, does not simply follow fixed commands. Instead, it can adapt to the environment around it and respond to obstacles as they appear. According to the findings shared by the cloud security firm Sysdig, the malware entered a system through a zero-day weakness in Langflow, an open-source AI development tool. It then accessed a Nacos server and a MySQL database, which helped the attack spread. In one case, the AI agent analysed a failed attempt to create a back door and solved it in 31 seconds by trying a different method. Experts say this could mark an important turn in ransomware, since key steps are now becoming fully automatic.