AI Stack Alert High TD-2026-0023 For CISO, SOC / IR, AppSec, AI Eng AI-assisted
LLMJacking: stolen LLM access costing victims over $100K per day, per Mandiant
Google Mandiant reports that attackers hijack corporate LLM access to run up token charges, sometimes exceeding $100,000 per day, with cumulative losses reaching $1 million at some organizations before access was revoked. Google documents a division of labor, with initial-access actors and separate "LLM Mixers" that monetize stolen access, and a service called Jet Night that allegedly relies on stolen keys.
Why it matters
Any organization with cloud LLM API keys or service accounts can incur large unbudgeted charges, and attackers gain cheap, high-volume compute for autonomous attacks. Stolen credentials can also bypass MFA when session tokens are captured.
Do this
- Inventory all LLM API keys and cloud service accounts with model access, and set spending alerts and quotas on each project.
- Rotate keys and secrets through a centralized secrets management platform, and revoke any keys or service accounts that are unused or unexpected.
- Apply least-privilege permissions to users and service accounts so only the access required for their roles is granted.
- Run automated secrets scanning across codebases and remove any API keys found in source code.
- Hunt for unexpected key creation, newly enabled AI services, short-lived service account churn, and billing spikes in cloud audit logs.
Signals
- NoCISA KEV
- 37/100Priority
- 85%Confidence
Evidence
“These fraudulent charges from resource consumption can exceed $100,000 a day, according to a Monday report by Google's Mandiant incident responders”
“In one case detailed in the report, a group Google tracks as UNC6993 illicitly gained access to LLM API keys, then generated additional keys and enabled services within the victim’s cloud projects, racking up $100,000 in charges.”
“Google predicts miscreants will “increasingly prioritize LLMJacking” compared to other types of resource theft like cryptojacking.”
Sources
Affected / related
UNC6993, Google Mandiant, Google Threat Intelligence Group, Jet Night, LLMJacking, CrowdStrike
Published 2026-11-02 16:00 UTC · updated 2026-11-10 00:24 UTC · AI-assisted summary, checked against quoted sources · How we score