Google Warns LLMjacking Attacks Cost Enterprises Up to $1M
Original title: Google Warns LLMjacking Attacks Are Costing Some Enterprises Up to $1 Million in Cloud Charges
Threat actors hijack LLM API keys to fraudulently consume expensive cloud resources with daily charges exceeding $100,000.
Google and Mandiant report attackers hijacking enterprise LLM access via stolen API keys, IAM credentials, or exposed tokens. They leverage the compromise to run proxy endpoints and agentic workloads, racking up bills billed to victims. UNC6993 and LLM Mixers are observed in related operations.
Unconfirmed: Specific models used in attacks not confirmed
- Fraudulent charges reached $100,000 USD in a single day
- Impacted sectors include finance, retail, technology, transportation
- Used stolen IAM credentials and GitHub PATs
- Objective is monetization through token proxying and GPU provisioning
Impact: unauthorized LLM resource consumption
- 1.revoke exposed API keys
- 2.monitor cloud billing
- 3.remove default passwords
- CVEs
- CVE-2026-40217
- Products
- LLM services
- Threat actors
- UNC6993
- Victims
- enterprise organizations
Vulnerabilities mentioned
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 | 16% | PoC |
From the original article
6 min read at thehackernews.com
Google on Monday warned that threat actors are hijacking enterprise access to Large Language Models (LLMs) to fraudulently leverage expensive computing resources and AI services. The tactic is called LLMjacking, a form of resource hijacking adapted for the LLM era, where an attacker obtains an API key through phishing, source code leaks, misconfigured environments, or exposed client-side code
thehackernews.com
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.