CVE-2026-40217
CVSS 8.8 HIGH: liteLLM through 2026-04-08 allows remote attackers to execute arbitrary code via bytecode rewriting at the… EPSS 16% (97th percentile).
Threats & Adversaries · Credential Theft
Google Mandiant says it responded to multiple LLMJacking incidents in 2026 across four sectors and three regions, with fraudulent usage exceeding $100,000 in a day and reaching $1 million cumulatively at some organizations. The cases affected financial services, retail, technology, and transportation organizations in North America, South America, and Asia Pacific.
In one case, the group UNC6993 used stolen LLM API keys to create more keys and enable services in a victim’s cloud project. Mandiant also describes operators called “LLM Mixers” that proxy and resell unauthorized model access, separating initial break-ins from downstream sales. A stolen key can therefore lead to both expanded access and a market for the stolen capacity.
For organizations whose cloud credentials can create keys or enable paid services, revoking the first token may leave attacker-created access—and its billing path—inside the project.
2 sources · 6h ago
CVSS 8.8 HIGH: liteLLM through 2026-04-08 allows remote attackers to execute arbitrary code via bytecode rewriting at the… EPSS 16% (97th percentile).
The Register Security
Token thieves costing companies $100K+ per day in illicit LLM charges
Don't let the LLMjackers hit you
originalMandiant
The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources | Google Cloud Blog
Mandiant reveals how LLMJacking attacks cost victims hundreds of thousands in fraudulent cloud bills, with underground markets targeting LLM credentials.
originalPart of the PlainSec briefing for 2026-11-02
Every edition of this story: Stolen LLM keys now feed a resale economy