Cybercriminals are building a business around reselling access to large language models (LLMs) through cloud resources, according to the reported findings. The activity, described as “LLMjacking,” can leave victims responsible for substantial consumption charges. In some cases, the bill reportedly reached as much as $1 million before access was revoked.
The central issue is the misuse and resale of cloud-based access: an unauthorized party uses resources associated with a victim and makes that access available to others. The reported figure illustrates the potential financial exposure, but no details were provided about the affected organizations, how their access was obtained, the number of incidents, or the period over which the charges accumulated. The available information also does not identify particular cloud providers or LLM services.
Why resold access creates a costly exposure
Cloud-hosted model usage can consume billable resources. When access is abused at scale or used by multiple customers of an illicit reseller, the account holder may face consumption costs it did not authorize. In the reported cases, access was eventually revoked, but only after costs had reportedly climbed as high as $1 million. Revocation can stop further use of the affected credentials or resource, though it does not by itself explain how the intrusion occurred or resolve charges already incurred.
The episode highlights an important distinction for cloud security teams: protecting access is not only about preventing data exposure. Unauthorized use of cloud services can also create direct financial consequences. For organizations that make LLM capabilities available through cloud accounts, effective oversight needs to include who can use the resources and whether usage matches expected activity.
Controls teams can review
The limited information does not establish which specific safeguards were absent in the reported incidents. Organizations can nevertheless use the cases as a prompt to review their own access and consumption controls. Measures to consider include restricting credentials to necessary users and workloads, revoking access that is no longer required, and monitoring usage and spending for unexpected changes. Teams should also know how to disable affected access promptly and investigate which credentials and resources may have been exposed.
Cloud access security brokers (CASBs) and other cloud security controls may help organizations apply access policies and gain visibility into activity, depending on their environment and configuration. They are not, on their own, evidence that a particular account is protected; teams need to verify that relevant services, identities, and usage are covered by their controls. Billing alerts and usage limits, where available, can complement access protections by making unusual consumption easier to detect.
LLMjacking’s reported resale model makes cloud access a commodity for criminals and places model consumption costs among the risks of compromised accounts. The available account is brief, so it does not support conclusions about the scale of the trade or the most common route to access. Still, the reported million-dollar exposure underscores the value of combining identity governance, usage monitoring, and a rapid response process for suspicious cloud activity.