NYU researchers develop AI-powered ransomware prototype
NYU researchers created PromptLock, a proof-of-concept ransomware that uses large language models to autonomously execute attacks through prompt-based instructions. The non-functional prototype was uploaded to VirusTotal during testing, where security researchers discovered it without knowing its academic origin.
What happened
A team of computer science professors and researchers at NYU's Tandon School of Engineering created PromptLock, a proof-of-concept ransomware prototype that uses an open-source large language model to autonomously plan and execute ransomware attacks.
The orchestrator component operates by delegating planning, decision-making, and payload generation to an LLM through natural language prompts, which then generates and executes malicious code dynamically. The researchers uploaded the non-functional prototype to VirusTotal during final testing without indicating its academic origin, leading cybersecurity vendor ESET to flag it as a discovered malware sample.
Reported impact
The research demonstrates how open-source large language models could enable new forms of ransomware attacks. The prototype can perform reconnaissance, payload generation, and personalized extortion in a closed-loop attack campaign without human involvement, creating polymorphic variants that adapt to the execution environment.
Index analysis
The researchers used prompt phrasing that makes requests appear legitimate rather than explicit jailbreaking techniques, allowing the LLM to receive only specific task instructions rather than the full attack orchestration. Some tasks such as data extraction and destruction faced refusals from the LLM despite these techniques.
Governance controls
The project was developed as a contained lab environment proof-of-concept that is non-functional outside that setting. The research was funded by grants from the Department of Energy, the National Science Foundation, and New York's Empire State Development's Division of Science, Technology and Innovation.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.