PromptLock

PromptLock is a proof-of-concept AI-enabled ransomware prototype, also associated with the academic label Ransomware 3.0. It is written in Go and is notable for embedding a locally accessible large language model into its execution flow rather than relying solely on prebuilt malicious logic. During runtime, it uses a local Ollama-served gpt-oss-20b model to generate Lua scripts on demand, allowing the malware to adapt actions such as filesystem enumeration, file inspection, data selection, exfiltration, and encryption to the victim environment. Reported code also contains indications of unfinished destructive or wipe-oriented functionality.

PromptLock has been described as one of the earliest concrete examples of ransomware using generative AI dynamically during execution. Its generated Lua logic has been reported as cross-platform, with observed variants for Windows and Linux and references to compatibility extending across additional desktop operating systems. The malware’s behavior demonstrates how publicly available local AI tooling can be incorporated into ransomware to increase flexibility and reduce the need to hardcode all malicious functionality into the binary.

Available reporting indicates PromptLock was not observed in real-world criminal deployment and is best assessed as a laboratory prototype or work in progress rather than an operational ransomware family used in active attacks. Multiple accounts link it to controlled research by New York University’s Tandon School of Engineering. Even so, it remains significant as an early demonstration of LLM-orchestrated ransomware design and of how AI can be used to automate reconnaissance, target-file selection, exfiltration decisions, and encryption workflows inside malware.