Malware uses LLMs to rank victims by how much they're worth (blog.talosintelligence.com)

A new malware dubbed VOLTA has emerged, integrating advanced AI capabilities to enhance its functionality as a stealthy information stealer. After pilfering valuable data such as credentials and cryptocurrency wallets, VOLTA employs a large language model (LLM) to generate a detailed analysis of the victim's assets and their worth. This analysis ranks the victim's valuable items by monetization potential while assessing their security posture, thus relieving the operator from tedious analytical work. Remarkably, the malware operates entirely within the victim's network, exfiltrating the resulting AI-generated dossier without relying on attacker-controlled infrastructure. The significance of VOLTA lies in its novel approach to cybercrime, employing AI to sift through vast amounts of stolen data and create tailored victim profiles. This represents a shift from traditional malware techniques, highlighting the growing trend of leveraging AI for sophisticated cybercriminal operations. The roadmap for defending against such threats includes restricting app behavior to combat techniques that bypass security measures and enhancing credential storage practices. As AI integrates deeper into the malware landscape, this advancement may lower operational barriers for cybercriminals and transform the complexities of cyberattacks.