Criminal Market Thrives on Reselling Cloud AI Access, Some Victims Lose Up to $1 Million

Law‑enforcement and cybersecurity analysts have identified a growing underground market that trades in rented access to large‑language‑model (LLM) services hosted in the cloud. Dubbed “LLMJacking,” the practice involves criminals hijacking legitimate cloud accounts or purchasing compromised API credentials, then reselling the compute power to other bad actors. The model mirrors earlier ransomware‑as‑a‑service schemes, turning the provision of AI resources into a repeatable revenue stream for organized cybercrime groups.

The mechanics are straightforward: once a threat actor gains control of a cloud subscription—often through phishing, credential stuffing, or exploiting misconfigured services—they can generate unlimited text, code, or data‑analysis calls using the provider’s LLM APIs. They then market this capability on underground forums, promising high‑throughput access at rates that undercut official pricing. Because the underlying cloud bill is charged to the compromised account, the original subscriber bears the cost while the criminal pocket‑books the profit.

Investigations disclosed that some victims have accrued charges approaching $1 million before the unauthorized usage was detected and the access revoked. In those cases, the perpetrators ran continuous generation tasks, exploiting the pay‑as‑you‑go billing model to inflate expenses dramatically. The financial impact extends beyond the immediate bill; organizations must also allocate resources to forensic analysis, account recovery, and remediation of any data that may have been exfiltrated during the abuse.

Cloud providers are responding by tightening anomaly detection and introducing usage caps for high‑volume API calls. However, the rapid adoption of generative AI has outpaced many security controls, leaving a gap that criminals are eager to fill. Experts warn that as more enterprises integrate LLMs into production workflows, the attack surface widens, making it easier for threat actors to locate valuable, under‑monitored accounts.

The emergence of LLMJacking highlights a broader shift in cyber‑crime economics, where the commoditization of AI services creates new profit vectors. Law‑enforcement agencies are beginning to treat illicit AI usage as a distinct category of digital crime, coordinating with cloud vendors to trace suspicious API traffic. Nonetheless, attribution remains challenging because the illicit activity often routes through multiple compromised accounts and anonymizing services before reaching the final buyer.

Security professionals advise organizations to adopt multi‑factor authentication, regularly audit API key permissions, and implement real‑time monitoring of AI‑related spend. By establishing spending thresholds and alerts for atypical usage patterns, companies can reduce the window of exposure. As the market for rented AI compute matures, ongoing collaboration between industry, regulators, and investigators will be essential to curb a practice that threatens both financial stability and data confidentiality.