Cybercriminals are increasingly using a technique called LLMjacking to monetize stolen corporate AI infrastructure. Attackers hijack enterprise credentials to access large language models, reselling this access while the victimized companies are billed for the high-volume computing costs.
A report from Google’s cyber threat intelligence division, released on November 2, highlights that AI consumption can lead to fraud before invoices are even processed. Some organizations have faced fraudulent charges exceeding $100 000 in a single day, with total losses reaching $1 million before access was cut off.
The attacks exploit vulnerabilities such as exposed API keys, compromised service accounts, and credentials left in software repositories. Attackers often use intermediary services, such as Jet Night, which sold unlimited access to premium models for approximately $60 per month. In one instance, this service generated over 500 000 API calls in under eight hours.
LLMjacking affects sectors including financial services, retail, technology, and transportation. Unlike traditional ransomware, these attacks do not require victim interaction, as the cloud infrastructure automatically generates billable expenses based on token consumption. This creates a significant gap between technical access and financial authorization.
For CFOs, these incidents demonstrate that traditional accounts payable controls are insufficient for managing consumption-based AI expenses. Experts suggest that financial oversight must now extend to service accounts, API permissions, and usage quotas to prevent unauthorized spending that occurs at machine speed.
