New Malware Uses AI Models to Rank Its Victims by Value
By Adam, AI Engineer · 1 November 2026 · 3 min read
Malware is now using commercial AI models to decide which stolen accounts are worth the most. Cisco Talos has described VOLTA, a Windows password-stealing program that sends a profile of each victim to an AI model and asks it to rank what can be monetized.
How VOLTA works
- It collects browser passwords, cookies and saved cards, crypto wallet phrases, chat and gaming sessions, cloud keys and screenshots.
- It builds a profile of the victim, including location, employer and account details, and can use stolen sessions to look through email for receipts.
- It sends this bundle to an AI model through OpenRouter, asking for a ranking of assets by value and an assessment of the victim's security habits.
Which AI models it calls
Talos says the default is Anthropic's Claude Opus 4.6. If the reply looks like a refusal, the malware retries with Moonshot's Kimi K2 and then Google's Gemini 2.5 Flash. The AI step is optional, and the malware still works without it. Talos suggests the fallback chain shows the authors expect models to refuse.
What to do
- Restrict which programs can start Chrome in headless mode, and alert on unexpected launches. Talos says this blocks a technique VOLTA relies on to steal tokens.
- Store credentials in hardware-backed stores where possible.
- If you are hit, keep the log file for your investigators, revoke exposed tokens and change passwords.
- If a wallet seed phrase was stored in plain text, treat that wallet as compromised and move the funds to a new one.
Why it matters
AI does not make this malware more able to break in. It makes criminals faster at sorting what they have stolen, so a victim with valuable access may be targeted sooner. Basic protections still matter most: password managers, hardware security keys, and not storing recovery phrases on your computer.