Hackers are adding artificial intelligence capabilities to information-stealing malware, enabling it to profile victims automatically and identify valuable accounts and data more quickly, according to research from Cisco Talos shared with Kernel Panic. The malware, called “Volta-Autodoxer” because of how it refers to itself internally, uses commercial large language models (LLMs) to analyze stolen information and prioritize victims.
AI-assisted victim profiling
Cisco Talos researchers say the malware demonstrates how AI can augment established malware categories. Talos identified Volta using its CAIRN framework, which is designed to detect hacking tools that incorporate AI capabilities. However, Ryan Fetterman, a security researcher at Cisco Talos, said the company does not track whether Volta is currently being used.
“Volta is pretty mature as a stealer. It casts a pretty wide net in what it collects,” Fetterman said. “We compile all of this together and ultimately create a serialized document that can be sent to several LLM models. We ask the models to act like intelligence analysts and create structured victim assessments backed by citations. This is like a prioritization and analysis function.”
What Volta steals
Information stealers are often hidden in counterfeit software and are designed to capture usernames, passwords, authentication tokens, cookies, credit card data and other sensitive information. The malware category has been linked to major data thefts, including the massive Snowflake-related breaches involving AT&T and Ticketmaster, as well as the theft of terabytes of data. In one case involving Disney data, a hacker accessed the target using login details collected by an information stealer. Overall, this type of malware has been used to steal hundreds of millions of credentials in recent years.
Volta’s conventional infostealer capabilities appear particularly suited to younger internet users with extensive digital lives. It searches browsers for stored passwords, Microsoft OAuth tokens, cookies, payment cards and crypto wallet seed phrases. It can also access active sessions for Discord, Telegram, Steam and Roblox.
The malware uses the results of that initial collection to conduct a second, more detailed sweep. It examines Gmail and other central accounts, then uses information such as receipts and order histories from services including UberEats and DoorDash to build a broader profile of the victim, including their location and habits.
Multiple AI models and fallback systems
After collecting the data, Volta uses OpenRouter, a service that can route prompts to multiple AI providers, to submit the results for analysis. Its default request is sent to Anthropic’s Claude Opus 4.6. If the model’s guardrails block the activity, the malware includes a fallback to China’s indiscriminate weight AI’s Moonshot AI Kim K2, followed by Google’s Gemini 2.5 Flash if that option fails.
According to Talos, incorporating fallback models and relying on cost-effective legacy systems are common techniques among attackers. The approach allows malware operators to continue processing stolen data when one AI service refuses a request or becomes unavailable.
Source: www.wired.com