Drone-Bench: Tracking simple drone surveillance capabilities of frontier models
Summary
Drone-Bench is a benchmark from Andon Labs that measures how well AI models can write code to surveil real-world environments using low-cost drones. It defines five interrelated tasks—Reconstruct, Localize, Navigate, Detect, and Follow—each scored against a human baseline, with results tracked across frontier models. The piece discusses end-to-end implications, safety and misuse concerns, and future work to move beyond isolated tasks toward end-to-end evaluation.