Cybercriminals are increasingly targeting accounts linked to leading AI platforms, creating an underground market for unauthorized access to expensive large language models.
Ā
The technique, known as āLLM-jacking,ā uses stolen credentials or compromised infrastructure to access AI services without paying legitimate subscription and compute costs.Ā
Researchers report stolen account data can sell for a fraction of legitimate access costs, making compromised AI credentials attractive to cybercriminals seeking scalable computing resources.
Ā
Credentials are commonly harvested through infostealer malware, which compromises devices and captures browser data, saved passwords, financial information and GenAI account credentials.
Ā
Attackers can also exploit reverse-proxy techniques to route AI requests through compromised accounts.Ā
Ā
Sustained unauthorized consumption can potentially generate substantial bills for organizations whose credentials have been hijacked.
Ā
The threat extends beyond financial losses.Ā
Ā
Stolen AI access can support phishing, malware development, automated cyberattacks and potentially espionage while making malicious activity harder to attribute.
Ā
Organizations must therefore treat AI credentials as high-value digital assets, enforcing multi-factor authentication, securing endpoints, rotating credentials, monitoring abnormal AI consumption and immediately investigating unexpected usage spikes.















