Case promptlock-ransomwareAI-assisted malwareWorldwide
ESET finds PromptLock, ransomware that uses a local gpt-oss model to write scripts
ESET described PromptLock, a Go ransomware prototype that prompts gpt-oss:20b via Ollama to generate Lua scripts for discovery, exfiltration and encryption; likely a proof of concept.
LowAttack using AIStatus: ResolvedVerified case file
Open in the interactive mapCite this case
- Reported
- Aug 26, 2025
- Location
- Worldwide
- Target
- Windows and Linux systems
- Severity
- Low
- Triggered by
- NYU Tandon researchers (Ransomware 3.0 academic prototype)
- Actor type
- Security researchers
- Reported loss
- None reported
- CVE
- None
What happened
ESET described PromptLock, a Go ransomware prototype that prompts gpt-oss:20b via Ollama to generate Lua scripts for discovery, exfiltration and encryption; likely a proof of concept.
Impact: No real-world victims; destruction features unimplemented; demonstrated feasibility of LLM-driven ransomware across Windows and Linux
Who triggered it
Actor: NYU Tandon researchers (Ransomware 3.0 academic prototype)
Attribution: NYU Tandon team publicly claimed authorship after ESET's disclosure; ESET agreed the academic prototype closely resembles the PromptLock samples
Motive: Academic research into LLM-orchestrated ransomware; samples uploaded to VirusTotal to test detection
How it happened
Attack vector: Go binary with hard-coded prompts calling OpenAI gpt-oss:20b via the Ollama API to generate Lua scripts for file enumeration, exfiltration and encryption
Root cause: Freely available open-weight LLMs can be embedded in malware to generate polymorphic payloads at runtime; sample reached VirusTotal without research labelling
Timeline
- Windows and Linux PromptLock samples uploaded to VirusTotal
- ESET publishes analysis calling it the first AI-powered ransomware, judged a proof of concept
- NYU Tandon researchers reveal PromptLock is their Ransomware 3.0 prototype and non-functional outside a lab
Response
ESET published detections and IOCs; NYU clarified the research origin and stated the prototype was lab-only
Current status: Resolved
Sources (open in a new tab)
- Security Affairs securityaffairs.com
- The Hacker News thehackernews.com
- Security Boulevard securityboulevard.com
- Dark Reading darkreading.com
Cite this case
ESET finds PromptLock, ransomware that uses a local gpt-oss model to write scripts. AI Threat Watch, case promptlock-ransomware, Aug 26, 2025. https://aithreat.io/cases/promptlock-ransomware/
Permanent link: https://aithreat.io/cases/promptlock-ransomware/