Skip to content
Source: ZDNET

🔐Criminals Resell Stolen AI Access Through LLM Mixers, Some Batches Sell for $0.10

Your API keys are being resold, and you pay the bill

TL;DR

Google's threat team says criminals are reselling stolen enterprise AI access through storefronts called LLM Mixers. Token batches sell for as little as $0.10, while victims can face bills above $100,000. Lock down your API keys before someone else runs up your tab.

Criminals are now reselling stolen enterprise AI access like it's a commodity. Google's Threat Intelligence Group calls these storefronts LLM Mixers, and they've seen a major increase in this kind of LLMjacking throughout 2026. If your team uses an LLM for anything, from calendar invites to data analysis to code, this hits you. The attacker doesn't need your data to hurt you. They just need your key. Once they have it, they can run heavy compute, stand up their own malicious deployments, pull data out of the model's memory, or poison training data. And you're the one on the hook for the usage bill. The prices are wild. Token batches go for as little as $0.10, and business-use tokens range from $0.10 to $25. One vendor, APIMart, reportedly jacks its price fivefold once a prompt crosses 100,000 tokens. Mandiant has logged illicit AI usage topping $100,000 in just 24 hours, and Sysdig estimates LLMjacking could run around $46,000 based on earlier models. The catch? These storefronts spread queries across multiple victim environments to stay under the radar, so your spike may look like normal traffic until the invoice lands.

Criminals Resell Stolen AI Access Through LLM Mixers, Some Batches Sell for $0.10 — ZDNET

Key Points

1

Google's Threat Intelligence Group reports a major increase in LLMjacking throughout 2026, with stolen AI credentials now treated as a lucrative illegal business in their own right.

Keep reading — it's free

Enter your email to keep reading — plus the best of AI & tech, daily. Free, forever.

Also get
or

Already a member? Sign in

LLMjackingLLM MixersAPI keyssecrets managementAI security

Comments

Subscribe to join the conversation...

Be the first to comment