[xnx]

s/try/train/

[irobeth]

seconding this, it failed to correctly identify an AI image which had the DALL-E watermark clearly visible

what plans are there to guard against people intentionally poisoning your training data by miscategorizing the images they upload for classification?