LLMJacking: Criminals Hijack Corporate AI Resources for Massive Financial Gain
Unauthorized access to LLM API keys is costing organizations over $100,000 daily in fraudulent charges.
1The Rise of LLMJacking
Criminals are increasingly hijacking corporate LLM access to steal expensive computing resources. This tactic, known as LLMJacking, has grown steadily over the last two years. Victims face massive financial losses, with some organizations incurring over $100,000 in daily fraudulent charges before detecting the unauthorized resource consumption and revoking the intruders' access.
2Sophisticated Monetization Strategies
Threat actors like the group UNC6993 employ advanced methods to monetize stolen access. They frequently subscribe victim accounts to pay-as-you-go agreements for premium foundation models. By utilizing a burn and rotate strategy, they create and delete short-lived service accounts, effectively bypassing standard rate limits while monitoring the financial impact through billing interfaces.
3High-Volume API Abuse
The scale of these attacks is significant, with one documented case showing 200,000 API requests executed in just two minutes. Another incident involved an intruder performing over 500,000 API calls in under eight hours. These stolen keys often power third-party services like Jet Night, which offers suspiciously low-cost unlimited token plans.
4Division of Criminal Labor
The LLMJacking ecosystem now features a clear division of labor among cybercriminals. Some actors focus exclusively on establishing initial access to compromised cloud environments, while others manage the downstream monetization infrastructure known as LLM Mixers. This specialization allows attackers to scale their operations and maximize profits from stolen corporate AI resources.
5Deceptive Session Hijacking
Attackers use sophisticated social engineering to bypass security controls like multifactor authentication. By creating fake application portals, they trick users into pasting raw JSON payloads containing active access tokens. These tokens are then posted to backend APIs, granting criminals direct access to the victim's LLM services without needing further authentication credentials.
6Shifting From Cryptojacking
Google researchers predict that criminals will increasingly prioritize LLMJacking over traditional resource theft methods like cryptojacking. The ability to leverage agentic tools and gain a scale advantage over defenders makes AI service hijacking a highly attractive target. This trend poses a broader threat, potentially fueling a deluge of cheap, autonomous cyberattacks globally.
7Defensive Security Priorities
Organizations must prioritize identity and access management to mitigate these risks. Security teams should implement centralized secrets management platforms to rotate keys regularly. Additionally, applying the principle of least-access privileges ensures that service accounts only possess the permissions strictly necessary for their specific roles, significantly reducing the potential blast radius of compromises.
8Automating Secret Protection
Beyond access controls, companies should automate secrets scanning within their codebases to identify and remove hardcoded API keys. Proactive monitoring of billing interfaces is essential for early detection of anomalies. By securing the supply chain and enforcing strict credential hygiene, organizations can protect themselves from runaway costs and unauthorized AI service usage.