VOLTA infostealer uses AI to rank victims by financial value

Cybersecurity researchers have disclosed a new Windows infostealer, VOLTA-AUTODOXER (VOLTA), that embeds an AI-driven dossier capability to rank victims by monetisable value. The malware harvests a wide range of data from Chromium-based browsers (passwords, cookies), Outlook profiles and Slack workspaces, payment cards, cryptocurrency wallet seeds, and various session tokens (Discord, Telegram, Steam, Roblox), plus Microsoft OAuth tokens, SSH and cloud-provider keys, Wi‑Fi profiles and screenshots.

Cisco Talos’ Ryan Fetterman described that VOLTA “builds a structured evidence bundle about the person behind the machine,” listing items such as geolocation, home address, employer and domain, inferred birth year, card last-4 digits, Slack workspaces, document authorship metadata, iPhone backup indicators and a live Roblox account valuation.

The attacker stores the harvested data as a bundle and feeds it to an AI analysis layer via OpenRouter, using a system prompt that positions the AI as a forensic intelligence analyst. The prompt restricts facts to the structured bundle and instructs the model to produce sections titled WHAT THEY'RE WORTH (ranked low/medium/high) and THEIR OPSEC (security posture and password hygiene).

VOLTA’s AI chain attempts multiple model backends if a model refuses to answer, with fallbacks to Anthropic Claude Opus 4.6, Moonshot AI Kimi K2 and Google Gemini 2.5 Flash. Data exfiltration occurs through Telegram at three stages: immediately after execution, after the AI analysis, and after enrichment (Gmail receipts and order histories). A build-time template also stores API keys and the Telegram channel, suggesting the author may be selling a builder or operating a bespoke service.

Evidence and context from Talos also note related AI-enabled tools, including NIK C2, which can integrate an OpenAI client to process stolen data in-session, raising concerns about LLMjacking and attribution. The disclosure frames GenAI-enabled theft as a growing trend in threat actor workflows. Responders should monitor for similar multi-model AI-enabled stealers, assess Telegram exfiltration channels, and review access to Google and Gmail receipts that could corroborate victim addresses. The article cites Talos’ appraisal of VOLTA and situates the finding within ongoing use of GenAI in cybercrime. UK date: 2 November 2026.