Glonce

Volta infostealer sends stolen data to LLMs to profile victims, Cisco Talos finds

securityThe Glonce editorsWired🔥66Glonce rating66Impact55Novelty65Relevance85Surprise60Players35Credibility80
Volta infostealer sends stolen data to LLMs to profile victims, Cisco Talos finds

Cisco Talos researchers have found a new strain of infostealer malware that adds AI to the usual data grab. The finding was shared exclusively with WIRED's Kernel Panic newsletter, written by Lily Hay Newman and Matt Burgess. The malware is dubbed "Volta-Autodoxer" because of how it refers to itself internally. Its AI step lets it automatically profile victims and quickly surface the highest-value targets and their information.

Infostealers are usually hidden in counterfeit software and grab usernames, passwords, authentication tokens, cookies, credit card data and more. WIRED notes that many of the biggest data breaches in recent years, from the Snowflake-linked breaches that included AT&T and Ticketmaster to the theft of a terabyte of Disney data, resulted from hackers using login details gathered by infostealers. Overall, the malware has been used to steal hundreds of millions of credentials in recent years.

Volta's classic stealing is broad. It sweeps for passwords stored in browsers, Microsoft OAuth tokens, cookies, payment cards and crypto wallet seed phrases, plus active Discord, Telegram, Steam and Roblox sessions. WIRED says these capabilities seem to be geared toward younger internet users with robust digital lives. Then it does a second, deeper sweep using its first-round findings: by looking through the target's Gmail or another central account, it builds a fuller picture of the victim, their location and their habits, using receipts and order history from services including UberEats and DoorDash.

Next, the malware compiles everything into a serialized dossier and sends it to AI tools through OpenRouter, a service that can route prompts to multiple AI providers. Volta defaults to Anthropic's Claude Opus 4.6. Apparently anticipating that a model's guardrails could block the activity, it has fallbacks to the Chinese open-weight Moonshot AI Kimi K2 and, if that fails, to Google's Gemini 2.5 Flash. WIRED says building in fallbacks and relying on cost-effective, older-generation models are common attacker techniques.

The creators wrote a 240-word prompt, with bullet points and a numbered list, for whichever LLM analyzes the data. It opens: "You are a forensic intelligence analyst drafting a target assessment from a structured evidence bundle." It asks for an "executive summary" of the person, details of what they do, "what they're worth," and a summary of "their opsec." It tells the model to "connect dots across fields," for example by checking whether a Wi-Fi password matches the person's last name, and ends with "Go deep, not wide."

Ryan Fetterman, a security researcher at Cisco Talos, says the prompt is carefully crafted not to imply anything overtly malicious, yet the authors expect to hit model refusals at first request and plan a cascade to more permissive models. The implant also works without the AI: in his words, the AI calls are above and beyond. Talos flagged the malware with its new CAIRN framework, built to identify hacking tools that incorporate AI capabilities. Fetterman notes that Talos did not track whether Volta has been used in the wild. He says Volta contains some clues that it may have been built to sell to a range of attackers rather than for one actor's campaigns.

WIRED places the find in a wider trend. Some low-level cybercriminals complain about AI slop on cybercrime forums, but researchers increasingly find hackers using LLMs for coding or for analyzing troves of stolen data. Researchers at threat intelligence firm Crowdstrike recently told WIRED they had seen four distinct hacking groups of varying sophistication, operating in Chinese, Spanish and English, Indonesian, and Russian, using AI tools to compromise dozens of targets worldwide, including government entities and businesses in health care, retail and travel. Fetterman's argument for why Volta matters: faster identification of high-value targets raises the odds of using stolen credentials or tokens before they expire or passwords change.

Key facts

  • Cisco Talos found an infostealer dubbed Volta-Autodoxer that packs stolen data into a dossier and sends it to commercial LLMs via OpenRouter to profile the victim.
  • Volta defaults to Claude Opus 4.6, with fallbacks to Moonshot AI Kimi K2 and then Gemini 2.5 Flash in case guardrails block the request.
  • Its 240-word prompt casts the model as a forensic intelligence analyst and asks for an executive summary, what the victim is worth, and their opsec.
  • The malware still works without the AI step; Talos did not track whether Volta has been used in the wild.
  • Talos flagged it with its new CAIRN framework; Fetterman says it may have been built to sell to a range of attackers.

Why it matters

Infostealers already feed many large breaches, because stolen logins give attackers a way in. Volta adds a triage layer: instead of a human sorting through a pile of logs, an LLM reads the compiled dossier and ranks who is worth attacking. Fetterman describes it as a prioritization and analysis function and says it shows more advanced capabilities becoming accessible to lower-sophistication attackers, which he calls interesting and disturbing. WIRED says the malware at its core is still a traditional infostealer, and the AI expansion foreshadows more invasive and sinister activity ahead.

Who it affects

Potential victims are people whose machines get infected, apparently by way of counterfeit software. WIRED says Volta's feature set seems geared toward younger internet users with robust digital lives: it targets Discord, Telegram, Steam and Roblox sessions as well as browser passwords, Microsoft OAuth tokens, cookies, payment cards, crypto wallet seed phrases, and Gmail plus delivery-order history. It also matters to AI providers whose models sit in the fallback chain: Anthropic, Moonshot AI and Google.

How to use it

This is a threat report, not a product, so there is nothing to adopt. For security teams, the useful part is the detail: what Volta collects, the OpenRouter-based model chain, and the shape of the prompt. Talos found it with its new CAIRN framework, built to identify hacking tools that incorporate AI capabilities. Fetterman also points to time pressure: authentication tokens expire and passwords can be changed, which is why attackers want to find high-value targets fast.

How solid is it

The findings come from Cisco Talos researchers and were shared exclusively with WIRED's Kernel Panic newsletter. Details of the code and the prompt are described by WIRED and Fetterman directly. Talos did not track whether Volta has been used in the wild, so the research describes capability, not confirmed attacks. The idea that Volta targets younger users is WIRED's reading of the feature set ("seem to be"), and the suggestion that it was built for sale to many attackers is Fetterman's, based on "some clues".

Risks and caveats

No victims, creators or distribution details are named in the piece, and it does not say any AI provider responded or whether any model refused or complied with Volta's requests. The AI step is optional: the malware works without it. The Crowdstrike observation of four hacking groups using AI tools is a separate finding and is not tied to Volta. WIRED also notes that fallbacks and cheaper older models are already common attacker practice, so Volta's design is not unusual in that respect.

“We’re seeing more advanced capabilities accessible to lower-level sophistication attackers. It’s really interesting and disturbing.”

— Ryan Fetterman, security researcher at Cisco Talos