Researchers document an end-to-end agentic ransomware attack

The Register reports that Sysdig threat researchers documented what they describe as the first end-to-end agentic ransomware attack driven by an LLM. The incident, named JadePuffer, exploited an exposed Langflow instance, scanned for secrets, connected to a production MySQL and Nacos environment, created a backdoor administrator, encrypted configuration data, and generated a ransom note. Researchers warned that the attack lowered the skill floor for ransomware because the agent chained common weaknesses into a complete operation.

Key details: Sysdig named the agentic intruder JadePuffer, The attack exploited CVE-2025-3248 in an exposed Langflow instance, The LLM-generated operation encrypted 1,342 Nacos service configuration items.

Why it matters: The attack shows AI agents being used to automate the full intrusion chain, not just assist individual hacking steps.