Cybercriminals exploit API keys (authentication tokens) used by developers to access AI platforms through a technique called token jacking. By stealing these tokens via information stealers, phishing campaigns, compromised code repositories, and poisoned npm packages, attackers gain unauthorized access to expensive AI resources. These resources are either consumed by the attackers or resold through intermediary gray-market services known as transfer stations, which offer discounted access to frontier AI models using stolen credentials. The financial impact on victims can be catastrophic, with losses potentially reaching hundreds of thousands to millions of dollars before detection, due to default unlimited scaling and cyclical billing models. Mitigation strategies include enforcing spending limits, conducting privileged account reviews, using short-lived bearer tokens, deploying AI gateways to control access, and maintaining strict controls over development environments.

Unauthorized use of AI platform resources through stolen API tokens can lead to significant financial losses for organizations, potentially in the hundreds of thousands to millions of dollars. The exploitation leverages default unlimited scaling and cyclical billing, allowing attackers to incur large charges before detection. Additionally, stolen tokens are monetized in gray-market transfer stations, facilitating broader abuse of AI services. The compromise vectors include credential theft via malware, phishing, supply chain attacks (e.g., poisoned npm packages), and compromised code repositories.

There is no direct patch for token jacking as it exploits stolen credentials rather than a software vulnerability. Organizations should implement spending limits on AI resource usage to prevent runaway costs. Privileged account reviews should be conducted regularly to detect unauthorized token usage. Use short-term bearer tokens to limit the window of token misuse. Deploy AI gateways or access control mechanisms to monitor and restrict API usage. Maintain strict management and security of development environments to reduce the risk of token theft. These mitigations align with vendor and industry recommendations for protecting API keys and managing AI platform access.