Why did I get a wrong result on Reality Defender, and how should I interpret that?

If the model confidence was low (around 45-65% confidence), then it’s the domain where the neural network is not very sure. While if the model confidence is high for manipulation (above 70%), it can suggest that the result is a false positive. Conversely if the model confidence is low for manipulation (below 45% for most modalities), it could suggest the result maybe a false negative.