[danielrmay]

I've experienced this too - it's as if the security classifiers aren't keeping up with model intelligence. I'll leave the implication of that to the reader.