An inference API is a programmatic interface that lets an application send prompts or other input to an AI model and receive its output. It commonly handles requests to hosted models, with access controlled by credentials and cloud permissions.
In cyber security, an inference API is both a resource to protect and a point to monitor. If an attacker steals an API key or compromises a cloud identity authorized to use the service, they may submit unauthorized requests and generate usage-based charges. Defenders can reduce this risk with least-privilege access, protected and short-lived credentials, limits on model access, and alerts for unusual request volume, service activation, or spending.