Their philosophy is that false positives are much worse than false negatives (where "positive" means AI generated), so one would expect high precision and only

Their philosophy is that false positives are much worse than false negatives (where "positive" means AI generated), so one would expect high precision and only medium recall.