Security researchers at Cisco Talos have identified a new strain of malware that integrates artificial intelligence to automatically profile victims and identify high-value targets. Dubbed “Volta-Autodoxer” based on its internal naming conventions, the infostealer represents a shift in how cybercriminals use commercial large language models to enhance traditional data theft operations.
Automated Victim Profiling
Volta-Autodoxer functions as a standard infostealer by collecting usernames, passwords, authentication tokens, cookies, and credit card information from infected devices. However, it goes further by compiling this data into a serialized dossier and sending it to AI models for analysis. The malware instructs the AI to act as an intelligence analyst, producing a structured assessment of the victim.
“Volta is pretty mature as a stealer. It casts a pretty wide net for what it collects,” said Ryan Fetterman, a security researcher at Cisco Talos. “It compiles all of those things together, ultimately producing a serialized dossier that it can send to a couple LLM models. It’s asking the model to act like an intelligence analyst and produce structured citation-backed assessment of the victim. It’s like a prioritization and analysis function.”
The malware targets users with active digital lives, sweeping for browser passwords, Microsoft OAuth tokens, and session data from platforms like Discord, Telegram, Steam, and Roblox. It then performs a deeper sweep by accessing central accounts like Gmail to gather location data and habits from services such as UberEats and DoorDash.
AI Fallback Mechanisms
To process the stolen data, Volta-Autodoxer uses OpenRouter, a service that routes prompts to multiple AI providers. The malware defaults to Anthropic’s Claude Opus 4.6 but includes fallback options to Chinese open-weight model Moonshot AI Kimi K2 and Google’s Gemini 2.5 Flash. This cascading approach allows the malware to bypass potential guardrails that might block malicious requests.
The prompt sent to the AI is carefully crafted to avoid overtly malicious language. It instructs the model to draft a target assessment, including an executive summary of the victim’s worth and operational security practices. The prompt also asks the AI to connect disparate data points, such as checking if a Wi-Fi password matches a last name.
Fetterman noted that the prompt is designed to anticipate model refusals. “The prompt is pretty carefully crafted not to imply anything overtly malicious,” he said. “But they’re building in the expectation to potentially hit model refusals at first request. And so they plan for failover kind of cascading to more permissible models as they go down the chain.”
The malware can still operate without the AI component, but the integration allows for faster identification of valuable targets. This efficiency is critical because authentication tokens expire and passwords can be changed, giving attackers a limited window to exploit stolen credentials.
Broader Implications for Cybersecurity
The discovery highlights a growing trend of cybercriminals using AI tools to analyze stolen information and automate coding tasks. Threat intelligence firm Crowdstrike recently reported that multiple hacking groups operating in different languages are using AI to compromise targets in healthcare, retail, and travel sectors.
Cisco’s research suggests that Volta-Autodoxer may have been developed for sale to a range of attackers rather than for a single actor. This accessibility allows lower-sophistication criminals to perform tasks that previously required advanced expertise.
“If you’re able to identify your high-value targets faster, you probably have a better chance of using their credentials or authenticating using stolen tokens more often than maybe you would have if you had to do that triage yourself,” Fetterman said. “We’re seeing more advanced capabilities accessible to lower-level sophistication attackers. It’s really interesting and disturbing.”
Cisco Talos flagged the malware using its CAIRN framework, which is designed to identify hacking tools that incorporate AI capabilities. The researchers did not track whether Volta-Autodoxer has been used in widespread attacks yet.
Source: Wired
