🦠 Researchers Discover First “AI Computer Virus” Built on OpenAI Model
-
A new security report is making waves: ESET researchers say they’ve uncovered the first computer virus powered by AI, capable of writing malicious scripts in real time.
What They Found
The malware, dubbed PromptLock, uses an AI model (gpt-oss) whose weights OpenAI released as open source on Aug 5, 2025.
It generates Lua scripts on the fly, designed to work across Windows, Linux, and macOS.
Functions include:
Scanning device data.
Stealing or encrypting files.
A yet-to-be-activated data deletion function.
For now, researchers stress PromptLock is more of a concept / unfinished prototype than a fully functional virus.
Why It’s Significant
ESET calls PromptLock the first known malware built directly on AI models.
Unlike traditional malware, it can adapt and write new malicious code instantly.
Security analysts warn this could mark a new era of AI-assisted cyber threats.
Strange Detail: A Satoshi Wallet?
In one system prompt, researchers found a Bitcoin wallet address tied to Satoshi Nakamoto.
Cyber Security News notes this is likely just a placeholder or decoy, not a real connection — but it raises eyebrows.Big Picture
This discovery highlights the double-edged nature of open-source AI models:
They fuel innovation, but also open doors for malicious use.
Cybersecurity firms will now need to defend not just against static malware, but AI-driven, self-modifying code.
Question for the community: Do you think AI-powered malware will become a mainstream cyber threat, or is this still more of a proof-of-concept scare than a real danger?