PULSE the living trend engine
🤖 Open Intelligence Dossier available for AI agents & citation View Markdown (.md) →
◼ Archived Business

Spooky New Infostealer Malware Uses AI to ‘Autodox’ Victims

The VOLTA infostealer is leveraging LLMs like Claude and Gemini to automatically rank the financial value of its victims based on stolen data.

8sources
8articles
6velocity
+102%since first seen
15d agofirst detected

Velocity

How fast coverage is spreading — measured hourly from article rate × source diversity. How this works →

📍 How it ended

Coverage detailed how the new VOLTA infostealer malware uses AI models like Claude, Gemini, and Kimi to scan stolen data, assess victim assets, and rank targets by financial value. Reports from Cisco Talos and various publications outlined how the payload builds custom target assessments and can delete itself after execution.

The story quieted without a definitive conclusion in the coverage.

Epilogue added 13d ago, after coverage quieted.

The brief

A new piece of Windows malware known as VOLTA is utilizing artificial intelligence to profile and rank the financial value of its victims. According to reports from Cisco Talos and WIRED, the infostealer employs a process described as 'autodoxing' to assess targets. Once the malware infects a system, it raids the victim's Gmail receipts to gather data on assets. This stolen information is then fed into Large Language Models (LLMs), specifically naming Claude, Gemini, and Kimi AI, to determine the victim's worth. The payload builds a custom target assessment to help hackers prioritize which victims are most valuable for further exploitation. Coverage from Cisco Talos, The Hacker News, and Tom's Hardware emphasizes the technical flexibility of the VOLTA infostealer's AI stage.

If one AI model refuses to process the data or rank the victim, the malware is capable of switching to a different model to complete the task. Furthermore, Tom's Hardware notes that the malware can skip the AI-driven ranking step entirely if necessary. Mezha.net highlights a specific behavior where the malware may delete itself after it has tasked an AI model like Claude with ranking the valuables of the infected user, potentially to evade detection or clean up its footprint. This development represents a shift in how infostealers operate by integrating real-time appraisal tools into the attack chain. As detailed by The Indian Express and Martin Cid Magazine, the ability to profile victims in seconds allows attackers to move beyond bulk data theft toward high-value target prioritization. The reliance on Gmail receipts as a primary data source indicates that the malware specifically seeks out financial footprints and purchase histories to calculate a victim's net worth.

This automated prioritization reduces the manual effort required by hackers to identify the most lucrative targets among thousands of infected machines. Future monitoring of this threat will likely focus on the evolving interactions between malware and LLM providers. Based on the current coverage from Unite.AI and other outlets, the primary point of interest is how VOLTA manages the API interactions with Claude, Gemini, and Kimi AI. Observers will be watching to see if the malware adapts its data-feeding methods to bypass AI safety filters that might otherwise refuse to rank stolen personal data. The capability for the software to self-delete after the appraisal stage suggests a strategic approach to stealth that may be refined in future iterations of the VOLTA codebase.

Synthesized by PULSE from the headlines below under a strict no-invention contract. ✓ fact-checked: all claims supported by sources Updated 13d ago.

Quick answers

Which AI models does VOLTA use?

The malware utilizes Claude, Gemini, and Kimi AI to rank the financial value of its victims.

How does VOLTA gather information about a victim's wealth?

The malware raids the victim's Gmail receipts to collect data used for the AI appraisal.

What happens if an AI model refuses to rank a victim?

The malware can either switch to a different AI model or skip the AI ranking stage entirely.

Coverage (8)

Topics

Related trends

\n \n \n \n \n \n \n