A malware that makes four AIs vote to decide its next move
La noticia se lee sola: apareció un malware de Windows, CLOSEDQUORUM , que en vez de recibir órdenes de un servidor de mando le pregunta a cuatro modelos de IA —DeepSeek, Qwen, Mistral y Gemini— qué hacer, y ejecuta la acción más votada: robar, inyectarse, persistir o moverse. Cuando gana "robar", vuelca las credenciales de Windows desde LSASS, se lleva las contraseñas de Chrome, Edge y Firefox,…
Cisco Talos discovered a Windows malware called CLOSEDQUORUM that uses four AI models - DeepSeek, Qwen, Mistral, and Gemini - to decide its actions, which include stealing credentials and cryptocurrency. However, experts believe it is a proof-of-concept that is not functional, with publicly available API keys and a Discord webhook.
The malware's use of AI models to vote on actions makes it slower, more expensive, and more detectable. The article concludes that the use of AI models for malicious purposes is not inherently good or bad, but rather depends on the context, and that a similar approach can be used for defensive purposes, such as reviewing security changes.
Written by urgent.news from Dev.to's report — not a translation of it. Machine-written — may contain errors; check the original before relying on it.
