One prompt can fake a broken item, so my refund desk gives AI no way to pay anyone
This article was created with the help of AI, working from the project's code and real test runs. The cover and the pictures of the cup are AI-generated on purpose. The screenshots come from a real run. A buyer says the cup arrived torn and sends a photo. If you run a small shop, you have two bad options. Refund it and pay for the fakes, or question it and insult the honest majority. Adding a…
A single AI prompt can create a fake damaged item, allowing fraudulent refund claims. This issue gained attention when a customer reported a torn paper cup and received a refund. Fraud detection firm Forter reported a 15% increase in AI-manipulated refund claims in 2025. To investigate, a small seller named Break or Fake was created to test the system on PayPal's sandbox platform.
Break or Fake uses two AI vision models to analyze photos of damaged goods. The models must unanimously agree that the damage is plausible, otherwise the claim is held for review by a human. Refunds are limited to $15 for uploaded photos and $60 for live photos. To prevent fraudsters from manipulating the system, Break or Fake adds an extra layer of verification by checking the camera used to capture the photo.
If the camera is a known virtual camera, the claim is automatically sent to a seller for review. Despite the AI's attempts to fool the system, Break or Fake successfully detected fraudulent claims and prevented unauthorized refunds.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
