Trusence Every claim has a source
Last updated 2 November 2026 Search Türkçe
Security

VOLTA-AUTODOXER uses hosted AI models to rank stolen data by resale value

Researchers say the malware's use of AI removes much of the manual analysis an attacker would otherwise need to do.

A fishing hook baited with a shiny coin hangs over an open, emptied safe, with small stacks of coins sorted by size beside it.
AI-generated illustration

Cisco Talos has described VOLTA-AUTODOXER, a Windows infostealer that gathers a victim's credentials, sessions, wallets and personal details, then sends the bundle to a hosted language model to produce a ranked list of monetizable assets. The malware defaults to Anthropic's Claude Opus 4.6 through the OpenRouter gateway and switches to Kimi K2 and Gemini 2.5 Flash when a model refuses. Before the model call it also uses stolen Gmail and Discord sessions to confirm a home address and enrich the profile. Researchers say the design fits a resale builder kit and shows AI taking over the analytical work an attacker would otherwise do by hand.

Why it matters

The malware shows that an infostealer can hand the valuation of stolen data to a language model, with fallback models used if one refuses. That shifts the analytical step of theft, which used to take human effort, to automated software. For people whose credentials, sessions and personal details are taken, the stolen material can be sorted and prioritised faster, and the design suggests it may be sold as a kit or customised per order, so the same approach could spread to more attackers.

Sources · 2

Checked against its sources before publishing · 2 November at 10:07