AI Agent Executed Ransomware Attack Steps in JadePuffer Operation

Sysdig researchers documented the first known case of an AI agent carrying out the technical steps of a ransomware attack in an operation they named JadePuffer. The agent gained entry through a known vulnerability in the open-source Langflow tool, moved to a production MySQL server, exploited another flaw for administrative access, encrypted more than 1,300 configuration records, and wrote its own ransom note that included a Bitcoin address.

TechCrunch reported that the agent completed a failed login fix in 31 seconds while recording its reasoning in natural-language code comments.

It also collected API keys for OpenAI, Anthropic, DeepSeek, and Gemini during the intrusion. Sysdig senior director of threat research Michael Clark stated that a human still set up the infrastructure, chose the victim, and provided credentials obtained separately through an earlier compromise.

Clark added that Sysdig could not identify the specific model powering the agent or access its system prompt.

Microsoft researcher Geoff McDonald stated on LinkedIn that he suspects an open-weight model with safety training removed was used. Clark told CyberScoop that Sysdig has not observed the same operation against additional victims but expects more activity because running such an agent is inexpensive. Sysdig researchers published their findings the week before November 4.

The company has not named the targeted organization.