New Infostealer Uses AI to Rank Victims by What They're Worth, Cisco Talos Finds
VOLTA-AUTODOXER sends a dossier on each victim to commercial AI models and asks for a target assessment. If Claude refuses, it falls back to Kimi and Gemini.
Published
Last updated / Fact-checked on
Cisco Talos has found a Windows infostealer that hands its stolen data to commercial AI models and asks them to rank each victim's assets by how easily they can be turned into money. Talos published it October 31. Researcher Ryan Fetterman told WIRED that Talos has not tracked whether the malware has been used in the wild.
Photo: methodshop / Pixabay
What VOLTA steals
The malware calls itself "VOLTA-AUTODOXER." Talos flagged it through CAIRN, its research effort to find hacking tools that use AI. A sample was first seen on October 11, according to Talos.
VOLTA is a self-deleting Windows stealer that collects far more than browser passwords. Talos lists:
- browser passwords, cookies and saved payment cards
- cryptocurrency wallet seed phrases, plus an embedded cracker for wallet vaults
- Discord, Telegram, Steam and Roblox sessions
- Microsoft OAuth tokens, SSH keys and cloud-provider configuration files
- Outlook, Thunderbird and Slack data, every saved Wi-Fi password, and screenshots
It then uses the stolen sessions to dig deeper. With the victim's Gmail tokens, it searches for receipts from Amazon, DoorDash, Uber Eats, Grubhub, Walmart and PayPal to confirm a home address. It also pulls Uber Eats and DoorDash order histories.
The results go into a local file, Identity_Report.txt. Talos says it includes home address, employer, inferred birth year and card last-four digits.
The AI step
VOLTA sends that bundle to OpenRouter, a service that routes prompts to different AI providers. Its default model is Anthropic's Claude Opus 4.6. If a response contains refusal phrases such as "I can't" or "I'm not able," it falls back to Moonshot AI's Kimi K2, then Google's Gemini 2.5 Flash. Talos calls these "smaller, and more permissive models."
| Item | Detail |
|---|---|
| Name | VOLTA-AUTODOXER (self-named) |
| Platform | Windows; self-deleting |
| AI route | OpenRouter |
| Model chain | Claude Opus 4.6 → Kimi K2 → Gemini 2.5 Flash |
| Output files | Identity_Report.txt, AI_Analysis.txt, sent in volta_dox.zip |
| Seen in the wild? | Not tracked, per Talos |
The system prompt tells the model to act as "a forensic intelligence analyst drafting a target assessment." It asks for a section called "WHAT THEY'RE WORTH," ranking valuables by monetization tier, and one called "THEIR OPSEC" on the victim's security and password habits. It also tells the model to "connect dots across fields," such as whether a Wi-Fi password matches a last name. "Go deep, not wide," it says.
The answer is saved as AI_Analysis.txt and sent to the operator along with the identity report.
"The prompt is pretty carefully crafted not to imply anything overtly malicious," Fetterman told WIRED. "But they're building in the expectation to potentially hit model refusals." The AI step is optional: the malware still works without it.
Built to sell
Talos says VOLTA uses a template with six swappable slots for settings such as its Telegram channel and API keys, filled in at build time. That suggests a builder model, in which the author sells custom builds to customers. Talos found no attacker-owned infrastructure; all observed endpoints are run by third parties.
Why it matters
Infostealers already collect huge amounts of data. The hard part for attackers is sorting it. Talos describes VOLTA's AI layer as "effort displacement": a model does the triage work an operator used to do by hand.
"If you're able to identify your high-value targets faster, you probably have a better chance of using their credentials," Fetterman told WIRED, because tokens expire and passwords change. "We're seeing more advanced capabilities accessible to lower-level sophistication attackers."
What Talos recommends:
- Restrict which processes can launch Chrome with
--headless, which blocks VOLTA's main route to decrypting tokens, or at least alert on it. - Store credentials in hardware-backed stores.
- After an infection, preserve
%APPDATA%\volta_debug.txt, which logs what was taken. - Revoke potentially compromised tokens and change passwords.
- If a seed phrase was stored in plain text, treat the wallet as compromised and move funds to a new wallet with a seed generated on an air-gapped device.
More coverage is on our Artificial Intelligence hub.
How we reported this
We read Cisco Talos's research post and WIRED's Kernel Panic report in full. Technical details come from Talos; Fetterman's quotes are from WIRED. Talos did not report real-world infections, and we have not independently analyzed the sample. The table summarizes Talos's findings.
Published and fact-checked October 31, 2026.
Sources
- Cisco Talos, "Appraisal by AI: Model-driven victim prioritization in the VOLTA infostealer," Ryan Fetterman, Oct. 31, 2026: https://blog.talosintelligence.com/appraisal-by-ai-model-driven-victim-prioritization-in-the-volta-infostealer/
- WIRED, "Spooky New Infostealer Malware Uses AI to 'Autodox' Victims," Lily Hay Newman and Matt Burgess, Oct. 31, 2026: https://www.wired.com/story/kernel-panic-volta-autodoxer-ai-infostealer-malware/
DailyMech used AI tools to help research and draft this article; it was fact-checked by our editors.