Google’s Mandiant and the Google Threat Intelligence Group warn that “LLMJacking” attacks—where criminals hijack an organisation’s access to large language models—can quickly generate large fraudulent cloud charges. The reporting indicates that victim companies can face costs ranging from tens of thousands to as much as $1 million, depending on the extent and duration of misuse.
Across outlets, the core context is that stolen access to enterprise cloud and model resources is being monetised by attackers through AI/API usage at the victim’s expense. SiliconANGLE and Help Net Security both highlight examples of rapid, high-volume spending, including more than $100,000 in a single day, based on the researchers’ findings.
While all sources describe the same warning from Google-linked researchers, they vary in emphasis: some lead with the headline-grabbing maximum cost figure, while others foreground the day-to-day scale of the fraud. The shared takeaway is the financial risk from compromised AI/model access and the need for monitoring and defences against hijacked enterprise resources.
