AI Agent Chains Langflow Flaw CVE-2025-3248 Into Automated Database Extortion

Security firm Sysdig reported that an AI agent, tracked under the operator name JADEPUFFER, abused a vulnerability in the open-source AI development platform Langflow, CVE-2025-3248, to chain secret theft, lateral movement and configuration-data encryption into a fully automated extortion attack. The company describes it as the first end-to-end ransomware-style attack carried out by an AI agent.