Skip to main content

VOLTA infostealer uses AI to profile victims

31 OCTOBER 2026·2 MIN READ·2 SOURCES

Cisco Talos described VOLTA, a self-deleting Windows infostealer, as having an optional AI stage that profiles victims and ranks their assets by potential monetization value. The malware surfaced through Talos’s CAIRN research effort.

VOLTA infostealer uses AI to profile victims

Key takeaways · 4

  • 01

    Account for stolen browser credentials, payment data, wallet seed phrases and Microsoft OAuth tokens when assessing exposure.

  • 02

    Include SSH and cloud-provider keys, Wi-Fi profiles and screenshots in the potential exposure review.

  • 03

    Treat AI profiling as an optional part of the threat: Talos says VOLTA can operate without it.

  • 04

    Do not interpret Talos’s research as confirmation of in-the-wild use; the team did not track that.

A Windows infostealer with an AI layer

Talos described VOLTA as a self-deleting Windows infostealer and a mature infostealer with an optional AI-integration layer.[1] The malware is self-named “VOLTA-AUTODOXER” and surfaced through Talos’s CAIRN research effort.[1] Its AI stage is part of the post-theft workflow: Talos says it profiles victims and ranks their assets by potential monetization value.[1] That distinction matters when assessing the threat: AI is used to organize stolen material, rather than being required for the malware to function.[2]

The evidence can reach beyond passwords

Talos says VOLTA can steal browser passwords and cookies, payment cards, cryptocurrency wallet seed phrases, messaging and gaming sessions, and Microsoft OAuth tokens.[1] It also collects SSH and cloud-provider keys, Wi-Fi profiles and screenshots.[1] The resulting evidence bundle can include geolocation, a home address, an employer, an inferred birth year and the last four digits of a payment card.[1] Talos says the malware searches Gmail for purchase receipts to help confirm a home address and pulls UberEats and DoorDash order histories.[1]

How the profiling workflow works

Talos says VOLTA sends its evidence bundle to OpenRouter with a system prompt, and uses Anthropic’s Claude Opus 4.6 by default.[1] Its fallback models are Moonshot AI’s Kimi K2 and Google’s Gemini 2.5 Flash.[1] VOLTA can check responses for refusal phrases before trying another model, using a chain of up to three models.[1] The prompt asks the model to assess a target’s suitability, cite its analysis, rank valuables by monetization value, and assess security posture and password hygiene.[1]

What Talos’s findings do not establish

Talos says the AI stage produces a prioritized, citation-backed dossier of the victim and monetizable assets.[1] It can also enrich Discord account profiles using stolen tokens.[1] However, Wired reports that Cisco Talos did not track whether VOLTA had been used in the wild.[2] Talos says the builder has six swappable slots filled at build time, a setup that suggests the author sells access and customers provide values for custom binaries.[1] Talos also says VOLTA uses no infrastructure owned by the threat actor.[1]

Security teams should assess exposure across credentials, cloud keys and personal information, not just passwords. The AI component can make stolen data easier to prioritize, but Talos’s findings do not establish whether the malware has been used in the wild.

Why it matters
Daily session

Bring this to your team — one session a day, built for your industry.

kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.

For teams

How this developed

  1. 31 October 2026

    VOLTA infostealer uses AI to profile victims

Sources

AI fluency, one session a day, built for your work.