xAIVerified interview challenge
Design abuse controls for a public AI free tier
Protect an AI service when one operator can exhaust capacity through thousands of free accounts.
Unlock with Full AccessThe challenge
A public AI service is about to expand free usage after a coordinated campaign exhausted limits across thousands of accounts. Legitimate launches and shared university or company networks can produce similar traffic patterns.
Design the abuse-control layer using the supplied traffic, identity and cost evidence. Explain how quotas, detection and responses work before model execution, and how the service should behave when a dependency fails or the evidence is uncertain.
What you'll practise
Coordinated activity
Combine available identity and behaviour signals without treating linkage alone as proof.
Capacity protection
Explain how usage reservations, concurrency and failure behaviour protect serving resources.
Proportionate responses
Choose enforcement and rollout decisions that account for legitimate users.
Write your reasoning and submit it for feedback.