Cybersecurity researchers have identified the first known ransomware family leveraging artificial intelligence for its local operations. Dubbed PromptLock, this malware demonstrates how AI can be integrated into traditional ransomware workflows, even if it is currently only a proof-of-concept (PoC).

Developed in GoLang and using OpenAI’s GPT-OSS:20b, an open-weight AI model, PromptLock dynamically generates attack scripts. These scripts, written in Lua, allow the malware to perform filesystem enumeration, file inspection, data exfiltration, and file encryption. Both Windows and Linux variants have been observed, and the Lua scripts are cross-platform compatible. For encryption, PromptLock employs the SPECK 128-bit algorithm.

ESET, the security firm that discovered the malware, explains that while PromptLock could potentially exfiltrate, encrypt, or destroy files, the destructive functionality is not yet fully implemented.

How PromptLock Works

PromptLock relies on hard-coded AI prompts to generate attack scripts on the fly. The AI model runs locally through the Ollama API, which requires the Ollama service to be active on the victim system. This makes full-scale attacks challenging in typical environments, as most systems lack the required resources.

The malware has been observed sending requests within the local network, likely interacting with either a locally running Ollama server or an internal proxy redirecting requests to an external AI service. Successful execution also depends on poor network segmentation and the absence of safeguards against automated AI prompts or outgoing traffic restrictions.

Implications for Cybersecurity

Although PromptLock is not yet deployed in the wild, its emergence signals a new frontier in cyber threats: ransomware that leverages AI to adapt operations in real time. According to ESET, raising awareness about such AI-powered threats is crucial for the cybersecurity community to prepare, research, and discuss mitigation strategies.

The development of PromptLock highlights how AI can automate parts of the attack lifecycle traditionally handled manually, potentially lowering the barrier for sophisticated attacks in the future. As AI models continue to evolve, the risk of AI-assisted malware may grow, making proactive defenses and preparedness essential.

Source: https://www.securityweek.com/promptlock-first-ai-powered-ransomware-emerges