Cybersecurity researchers at ESET have uncovered the first known AI-powered ransomware, signaling a dangerous evolution in cybercriminal tactics that could fundamentally change how organizations defend against attacks.
The malware, dubbed "PromptLock," leverages OpenAI's gpt-oss:20b model through the Ollama API to dynamically generate malicious Lua scripts in real-time. Unlike traditional ransomware with static code signatures, this AI-driven approach creates variable attack patterns that could severely complicate detection efforts.
"PromptLock leverages Lua scripts generated from hard-coded prompts to enumerate the local filesystem, inspect target files, exfiltrate selected data, and perform encryption," ESET researchers explained. The ransomware, written in the versatile Golang programming language, has variants targeting Windows and Linux systems.
What makes PromptLock particularly concerning is its adaptive capabilities. The AI generates custom ransom notes based on whether infected systems are personal computers, company servers, or critical infrastructure like power distribution controllers. This personalization could make attacks more convincing and harder to identify as automated threats.
While currently assessed as a proof-of-concept rather than actively deployed malware, PromptLock demonstrates how readily available AI tools are lowering barriers for cybercriminals. The malware uses SPECK 128-bit encryption and includes capabilities for data exfiltration, though data destruction features remain unimplemented.
"PromptLock uses Lua scripts generated by AI, which means that indicators of compromise may vary between executions," ESET noted. "This variability introduces challenges for detection and could significantly complicate threat identification."
The discovery comes amid growing concerns about AI misuse in cybercrime. Anthropic recently banned accounts after threat actors used Claude AI to develop advanced ransomware variants, while researchers have identified numerous prompt injection vulnerabilities in major AI platforms.
Organizations should prepare for this new threat landscape by implementing AI-aware security monitoring, maintaining robust backup systems, and ensuring incident response plans account for adaptive malware behaviors.
As AI becomes more accessible, the cybersecurity community must evolve detection methods to match increasingly sophisticated AI-powered attacks.