The Rise of LLMJacking: How Threat Actors Monetize Compromised Cloud and Model Resources

Why it matters

Mandiant and Google Threat Intelligence Group describe a criminal tactic they call LLMJacking, in which intruders hijack enterprise access to large language models and run up fraudulent cloud charges far faster than cryptojacking typically does. The incidents spanned multiple cloud and LLM providers and affected organizations across several regions and industries, including financial services, retail, technology, and transportation. Rather than abusing customer-facing chatbots, attackers target backend model resources directly, hunting for publicly exposed LLM endpoints or valid API keys. The dominant entry route, however, has been exposed or stolen cloud identity tokens, such as IAM credentials, Service Account keys, GitHub Personal Access Tokens, and active session tokens. In one case, attackers used a misconfigured Service Account key to reach an LLM API and set up a fake application portal that tricked users into surrendering raw access tokens, bypassing password and multifactor authentication. Once inside, actors escalate privileges, clone repositories, and hunt for hardcoded secrets. The report also documents a maturing underground economy in which initial-access brokers sell to LLM Mixers, illicit gateways that resell unauthorized model access to subscribers.