Google Threat Intelligence Group and Mandiant report that attackers are hijacking enterprise access to large language models, a tactic known as "LLMjacking". In incidents investigated in 2026, fraudulent API usage exceeded $100,000 in a single day at some victims, and cumulative charges reached $1 million before access was revoked.
According to the report, the intrusions spanned several cloud and LLM providers and hit organisations in North America, South America and Asia Pacific, across financial services, retail, technology and transportation. The attackers typically obtain credentials first; a separate group then resells the stolen access through so-called "LLM mixers", illicit gateways that proxy and resell unauthorised LLM capacity to underground subscribers.
What this means for users of LLM APIs: a leaked API key is not just a data risk. It can be turned into a large bill within hours. Rotate keys regularly, never leave them in code repositories or CI logs, set spending limits and alerts, and revoke unused credentials.
The report was published on 2 November 2026 by Google Cloud's threat intelligence blog.