How an AI-Generated Photo Slipped Through DoorDash’s System

[bonk]

Hobart speculated that the driver may have used a compromised account on a jailbroken phone, combined with access to historical delivery photos. DoorDash allows drivers to view images from prior deliveries, which could have enabled the attacker to recreate a convincing scene and generate an AI image that closely matched the customer’s front door.

In response, DoorDash confirmed that it investigated the incident, permanently removed the driver from the platform, and refunded the customer. A company spokesperson stated that DoorDash has “zero tolerance for fraud” and relies on both automated systems and human review to detect abuse — highlighting growing challenges platforms face as AI tools make deception easier to scale.

[madmax]

“zero tolerance” always sounds nice until you realize how easy this is to repeat