AnswersPublished17 sources
LLMJacking: How Stolen AI Access Ran Up $1 Million Cloud Bills
Google reported that LLMJacking—criminals hijacking enterprise LLM access for their own use or resale—caused more than $100,000 in fraudulent charges in a day, with some organizations reaching $1 million before access... The incidents spanned financial services, retail, technology and transportation organizations in...
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- Google reported that LLMJacking—criminals hijacking enterprise LLM access for their own use or resale—caused more than $100,000 in fraudulent charges in a day, with some organizations reaching $1 million before access...
- The incidents spanned financial services, retail, technology and transportation organizations in North America, South America and Asia Pacific.
- A resale service called Jet Night was linked to more than 500,000 model API calls in under eight hours; the available sources do not settle every technical detail of the reported LiteLLM vulnerability.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What did Google’s Mandiant and Google Threat Intelligence Group report on November 2, 2026, about LLMJacking incidents across financial serv. Article summary: Google’s Mandiant and Google Threat Intelligence Group reported on November 2, 2026, that criminals were hijacking enterprise LLM access and monetizing it at victims’ expense, producing fraudulent charges above $100,000 . Topic tags: general, documentation, general web. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fak
Google’s Mandiant and Google Threat Intelligence Group reported multiple 2026 incidents in which criminals hijacked enterprise access to large language models (LLMs), then used or resold that access. The resulting fraudulent resource charges exceeded $100,000 in a day in some cases; at some organizations, cumulative charges reached $1 million before access was revoked. 4
What LLMJacking means
LLMJacking is the hijacking of an organization’s access to LLMs and related computing resources. The reported activity centers on abusing enterprise cloud and model access—not simply manipulating conversations in a customer-facing chatbot. Criminals could run their own operations on a victim’s account or route other users’ model requests through compromised access. 4
Mandiant said the incidents it handled affected organizations in financial services, retail, technology and transportation across North America, South America and Asia Pacific, and involved a variety of cloud and LLM providers. 4
How attackers got access—and turned it into bills
Reported access routes included stolen cloud credentials, LLM API keys and access tokens. Other incident summaries identify IAM credentials, service account keys, GitHub personal access tokens and hijacked session tokens among the compromised credentials. 3
5
9
Once inside, attackers could use victims’ authorized access to run model workloads. Some then resold that access through intermediary services, shifting the cost of inference—the processing required to respond to model requests—onto the victim organization. 3
The scale could build quickly. Google said fraudulent charges exceeded $100,000 in a single day in some incidents, while cumulative charges reached $1 million at some organizations before access was cut off.
UNC6993 and the LLM resale market
Google-linked reporting describes UNC6993 gaining access to a cloud instance through a misconfigured service account key and using that foothold to reach an LLM API. 17 The case illustrates how a cloud identity weakness can become a route to unauthorized model use.
A separate part of the criminal ecosystem consists of so-called “LLM Mixers”: services that relay paying customers’ requests through compromised model access. Jet Night is one example. Reporting linked it to more than 500,000 API calls in less than eight hours and nearly $1 million in token costs imposed on one victim. 18
Sources also report that Jet Night advertised access for about $60 a month, compared with legitimate subscriptions priced at roughly $100–$250. 18
20 The low advertised price helps explain the resale model: buyers pay the intermediary, while the organization whose credentials were stolen may receive the bill.
A reported LiteLLM exploit, with a caveat
Google-linked coverage also described attackers exploiting CVE-2026-40217 in LiteLLM, a model gateway, to turn ordinary virtual keys into administrator-level access. 17 However, another supplied source flags a discrepancy between Google’s characterization and a LiteLLM advisory response. The available sources do not resolve that disagreement, so the precise technical description should be treated cautiously.
Why the costs can spike so quickly
Unlike cryptojacking, which Google’s researchers contrasted with LLMJacking as a more gradual drain on resources, LLMJacking can generate large fraudulent cloud-charge spikes within hours. The reported cases show why unauthorized model access is also a cost-control problem: a compromised credential can enable high-volume use before an organization notices and revokes it.
For organizations using cloud-hosted models, the incidents point to practical priorities: protect cloud identities and API credentials, review where tokens are exposed, and monitor model usage and charges for unexpected changes. These are defensive implications of the reported access routes and billing impact, not a complete list of recommendations attributed to Google. 3
5
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "LLMJacking: How Stolen AI Access Ran Up $1 Million Cloud Bills"?
Google reported that LLMJacking—criminals hijacking enterprise LLM access for their own use or resale—caused more than $100,000 in fraudulent charges in a day, with some organizations reaching $1 million before access...
What are the key points to validate first?
Google reported that LLMJacking—criminals hijacking enterprise LLM access for their own use or resale—caused more than $100,000 in fraudulent charges in a day, with some organizations reaching $1 million before access... The incidents spanned financial services, retail, technology and transportation organizations in North America, South America and Asia Pacific.
What should I do next in practice?
A resale service called Jet Night was linked to more than 500,000 model API calls in under eight hours; the available sources do not settle every technical detail of the reported LiteLLM vulnerability.
Sources
cloud.google.comThe Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources | Google Cloud Blog
mallory.aiLLMJacking Turns Stolen Cloud Credentials Into Six-Figure AI ...
pranavaraparla.comActive alerts — Vibe Coding · Security Issue Tracking
ai-threat.watchAttacks on AI systems: latest reports and research - AI Threat Watch