Tia N.List

CTI · reviewed · 2026-09-29

Infostealers Harvest Replayable AI Session Tokens

Full story & timeline

Actors

—

Malware / tools

LummaVidarStealCRedLineAcreedAtomic Stealer

Campaigns

—

CVEs

—

Victim sectors

technologyenergyfinancial serviceshealthcare

Geography

—

MITRE ATT&CK

Indicators

Observables tied to this case, from the curated feed. raw sidecar

Detections

Detection content derived from this case analysis. Status: experimental — not validated against live telemetry.

Sigma raw

title: Sign-In From Known AI Token Theft Login Source
id: 9f4b27e0-6d13-4a88-b2c5-71ae9d3f5c46
status: experimental
description: Detects successful authentication from source IPs that Unit 42 documented as malicious login origins used with stolen or exposed AI API keys and session tokens integrated into gray-market transfer stations. A hit means a stolen token is being replayed from attacker infrastructure — treat as LLMjacking in progress.
references:
  - https://unit42.paloaltonetworks.com/ai-token-jacking/
  - https://www.okta.com/blog/threat-intelligence/free_tokens_for_sale/
author: Tia N. List
date: 2026/09/29
tags:
  - attack.t1528
  - attack.t1539
  - attack.credential-access
logsource:
  category: authentication
detection:
  selection:
    src_ip:
      - 117.72.74.48
      - 207.246.106.162
      - 23.236.182.215
      - 95.214.112.26
  condition: selection
falsepositives:
  - Very low: these IPs were published as attacker login sources; expect false hits only if the ranges are recycled after takedown
  - Analyst or threat-hunting traffic from the indicator itself
level: high

References

Source on GitHub