- Published on
New VOLTA Infostealer Uses Claude, Gemini, and Kimi AI to Rank Victims by Financial Value
- Authors
- Name
- DeepHorus
Cybersecurity researchers at Cisco Talos have discovered a new Windows infostealer called VOLTA-AUTODOXER (or VOLTA) that uses large language models to rank stolen victim data by monetization value. The malware employs a multi-model AI system to generate prioritized dossiers. This marks an escalation in adversarial AI use.
Like other stealers, VOLTA harvests passwords, cookies, payment cards, cryptocurrency seed phrases, and session tokens from browsers, Outlook, Slack, Discord, Telegram, Steam, and Roblox. It also collects SSH keys, Wi-Fi profiles, and screenshots. It builds a structured evidence bundle about the victim.
VOLTA uses OpenRouter to access AI models, defaulting to Anthropic's Claude Opus 4.6, with fallbacks to Moonshot AI's Kimi K2 and Google's Gemini 2.5 Flash. A 240-word system prompt instructs the model to act as a forensic intelligence analyst. The output includes sections titled "WHAT THEY'RE WORTH" and "THEIR OPSEC."
To corroborate home addresses, VOLTA searches the victim's Gmail for receipts from Amazon, DoorDash, Uber Eats, Grubhub, Walmart, and PayPal. Data exfiltration occurs over Telegram in three stages: after initial steal, after AI analysis, and after live enrichment. The malware uses a template with six build-time slots, suggesting a builder kit or bespoke tool.
According to Talos researcher Ryan Fetterman, the bundle contains geolocation, employer, inferred birth year, card last-4 digits, Slack workspaces, document authorship metadata, and a live Roblox account valuation. This enables operators to prioritize high-value targets. The AI layer relieves operators of analytical work.
Talos said VOLTA is a mature stealer with a late-stage AI layer added for effort displacement, not speed or scale. If commoditized, the AI dossier capability allows customers of any sophistication to receive ranked target assessments in their Telegram inbox. The disclosure comes as adversarial adoption of generative AI continues to rise.
