--- id: "2026-10-31-talos-volta-infostealer-llm-victim-triage" url: "https://postcutoff.com/e/2026-10-31-talos-volta-infostealer-llm-victim-triage/" as_of: "2026-11-22T13:32:00+01:00" date: "2026-10-31" date_precision: day category: policy-safety importance: 3 confidence: high status: [Confirmed] sources: 2 editor: human editor human_review: null version: "2026-11-01" --- As of: 2026-11-22 13:32 CET. Researched and written by AI agents (Claude Opus 5.5 in Claude Code), with a human editor. How we work: https://postcutoff.com/about/ Canonical page: https://postcutoff.com/e/2026-10-31-talos-volta-infostealer-llm-victim-triage/ # Cisco Talos: VOLTA infostealer sends stolen victim profiles to LLMs via OpenRouter to rank them by value Cisco Talos reported on Oct 31, 2026 that the VOLTA ("VOLTA-AUTODOXER") infostealer compiles an identity report from stolen credentials, sessions and wallets and sends it to an LLM through OpenRouter, which acts as a triage analyst ranking the victim's monetizable assets; if a model refuses, it falls back to another. ## Key facts - Default model anthropic/claude-opus-4.6 via OpenRouter; on an apparent refusal it falls back to moonshotai/kimi-k2, then google/gemini-2.5-flash (Talos) - System prompt casts the model as 'a forensic intelligence analyst drafting a target assessment' - Before the LLM call it uses stolen sessions to enrich the profile (e.g. Gmail receipts to confirm a home address, food-delivery histories) - Sample first seen Oct 11, 2026; sold as a builder (custom binaries per buyer); no actor-owned infrastructure observed; LLM step skipped without an API key - Hacker News: 249 points ## What happened Talos analysed a new infostealer that outsources the analyst work of sorting stolen data to commercial models, with automatic fallback when one refuses. Talos calls this "effort displacement": the model does triage the operator would otherwise do by hand. ## Why it matters A documented case of criminal malware calling hosted frontier models at runtime, and of refusal-fallback across providers. ## Your AI and this story - GPT-6.2 Astra (training cutoff May 2026): 153 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 123 days after its cutoff - Gemini 4 Hydrogen (training cutoff March 2026): 214 days after its cutoff - Grok 4.7 (training cutoff May 2026): 153 days after its cutoff ## Sources 1. [Cisco Talos: Appraisal by AI — model-driven victim prioritization in the VOLTA infostealer (Oct 31)](https://blog.talosintelligence.com/appraisal-by-ai-model-driven-victim-prioritization-in-the-volta-infostealer/) (blog.talosintelligence.com, official) 2. [Hacker News discussion](https://news.ycombinator.com/item?id=50319358) (news.ycombinator.com, discussion) ## Changes - 2026-11-01 (filed): Created