What Is LLMjacking? The New AI Cybercrime Stealing Cloud AI Compute

Mar 5, 2026

LLMjacking is a new cybercrime where attackers steal access to cloud-hosted AI models and use them for free — while the victim pays the bill.

In this video, we break down what LLMjacking is, how attackers exploit compromised credentials and exposed APIs, and why security teams should treat AI infrastructure as a high-value attack target.

Discovered by the Sysdig Threat Research Team, LLMjacking is quickly becoming the AI-era equivalent of cryptojacking — except instead of mining cryptocurrency, attackers run expensive large language models (LLMs) at scale.

Because AI workloads blend into normal traffic, these attacks can be harder to detect than traditional resource abuse.

In this video you'll learn:

  • What LLMjacking is
  • How attackers steal AI model access
  • Why compromised API keys and tokens are the main entry point
  • How attackers rack up massive cloud AI bills
  • The link between cloud identity security and AI security
  • Practical steps to detect and prevent LLMjacking

If your organization is experimenting with generative AI, cloud AI APIs, or LLM infrastructure, understanding this threat is critical.

Protecting AI workloads means protecting identities, credentials, and runtime environments.

Chapters:

00:00 What is LLMjacking

00:26 How LLMjacking works

00:59 Why it's like cryptojacking

02:07 Steps to prevent LLMjacking

02:54 Importance of runtime visibility

03:24 Conclusion

#llmjacking #aicybersecurity #aisecurity #cloudai #llmsecurity #cryptojack #generativeai #aiinfrastructure #apikeys #aihacking