> When we ran the full dataset through the deep scan, it caught every single confirmed phishing site with zero false negatives. The tradeoff is that it flagged all 9 of the legitimate sites in our dataset as suspicious
Huh? Does this mean it just flagged everything as suspicious?