Google and Mandiant warn LLMjacking drives million-dollar cloud bills
First merged summary.
Attackers resell stolen enterprise LLM access, generating up to $100,000 a day and as much as $1 million in fraudulent cloud charges.
Google’s Threat Intelligence Group and Mandiant say financially motivated actors hijack enterprise access to commercial large language models and resell it through proxy services. Stolen IAM credentials, service account keys, GitHub tokens, session tokens, and LLM API keys fund proxy endpoints, agentic workloads, and GPU use billed to victims. Some organizations incurred more than $100,000 in a day, with cumulative costs reaching about $1 million, including nearly $1 million routed through jet-night.com. Targets include finance, retail, technology, and transportation firms in North America, South America, and Asia Pacific. Several outlets say attackers exploited LiteLLM CVE-2026-40217. Google advises least-privilege access and key rotation.
- Access via IAM credentials, service keys, GitHub PATs, session tokens, and LLM API keys
- Some victims incurred more than $100,000 in fraudulent charges in one day
- One case passed nearly $1 million in token costs through jet-night.com
- Cyber Security News reports over 500,000 LLM API calls in under eight hours
- Outlets say attackers exploited LiteLLM flaw CVE-2026-40217
- The Hacker News says UNC6993 and LLM Mixers appear in related operations
Timeline
dated events in the reporting · oldest first- Nov 2, 2026GTIG detailed LLMjacking in published research · Cyber Security News
What each outlet adds
beyond the shared facts- Dark ReadingContrasts sudden resource spikes with gradual cryptojacking
- The Hacker NewsNames UNC6993 and LLM Mixers in related operations
- BleepingComputerjet-night.com passed nearly $1 million in token costs
- Cyber Security NewsDates GTIG research Nov. 2; cites 500,000 calls in eight hours
- GBHackersNames LiteLLM /key/regenerate abuse and public GitHub key harvesting
- CVEs
- CVE-2026-40217
- Products
- Gemini APIBedrockLLM services
- Threat actors
- UNC6993
- AI models
- large language modelsLLM
Coverage
oldest first · one row per article · UTC- · 8d ago
Dark ReadingFirst report
LLMJacking Emerges as Thriving Cybercriminal Enterprise
Cybercriminals resell stolen cloud access to commercial LLMs, with victims charged up to $1 million in resources before access is revoked.
- · 8d ago
The Hacker News
Google Warns LLMjacking Attacks Cost Enterprises Up to $1M
Threat actors hijack LLM API keys to fraudulently consume expensive cloud resources with daily charges exceeding $100,000.
Vulnerabilities in this story
All| CVE | Vulnerability | CVSS | EPSS | Signals | Published |
|---|---|---|---|---|---|
| CVE-2026-40217 | LiteLLM through 2026-04-08 allows remote attackers to execute arbitrary code via bytecode rewriting at the /guardrails/test_custom_code URI | 8.8 | 17% | PoC |