Radek Musil
About meSkills
Services
My Projects
AI news
Contact me
Back to AI news

VOLTA malware asks AI how much your data is worth

October 31, 2026
Kybernetická bezpečnost a hackování
Foto: Sora Shimazaki • Pexels

The Windows infostealer VOLTA steals passwords and crypto wallets — then sends the victim's profile to a language model via OpenRouter, which rates what's worth monetizing.

How VOLTA works

Cisco Talos researchers described the self-propagating Windows infostealer VOLTA-AUTODOXER on October 31. After the classic theft of passwords, cookies, and crypto wallets comes a new step: the malware builds a victim profile and sends it through the OpenRouter API to a language model with instructions to act as a forensic analyst.

The model produces a target assessment — ranking stolen assets by monetary value and rating the victim's security habits. If one model refuses, the malware switches to a chain of three others. The operator no longer has to analyze anything manually; the AI prepares a ranking of the most profitable victims.

Why it's a watershed

Attackers have so far used AI mainly to write phishing. VOLTA raises the bar: AI becomes an operational part of the attack, automating the part that used to require a human analyst. Scaling cybercrime gets a new dimension — out of thousands of victims, the most lucrative ones are selected automatically.

Sources

  • Appraisal by AI: VOLTA infostealer (Cisco Talos)
  • VOLTA infostealer victim-ranking (Unite.AI)

Our take

This report should be pinned to every company's noticeboard. Attackers no longer dig manually — AI automatically picks whom it's worth extorting. The defense is still the same, just more urgent: MFA everywhere, password manager, separate crypto wallets, regular backups. And if you use OpenRouter or similar gateways, watch your logs — VOLTA shows that public AI APIs can be abused as a free analytics layer for malware. A security team that ignores this is fighting the last war.

More news

October 28, 2026

Google releases Gemini 4 Carbon: a frontier model for cybersecurity

Google's strongest model leads in finding vulnerabilities and fixing them. Access is initially limited to a handful of cyber defenders — a public release will follow later.

Read article
October 26, 2026

OpenAI releases GPT-6.1 Terra: a cheap model for computer work

The new model at $1 per million input tokens understands the screen, clicks through apps, and writes naturally. It costs a tenth of Astra and handles tasks 9× cheaper than the competition.

Read article
Back to AI news
Radek Musil|Product Engineer
Čeština·English
|Helper·Privacy Policy·Cookies·Terms of Use|

© 2026 Radek Musil. All rights reserved.