Bogus Data within <style> tags poisoning SA results

This seems to be a newer tactic, and a lot of email with content
poisoning seems to be slipping through our spam filters.  The reason
is all the "legitimate" content between <style> tags.  Most email apps
don't show the data between <style> tags and therefore goes ignored
and unseen by the recipient, but SA seems to be looking at it and
using it to poison the scoring system.

Here's an example of what we're seeing within the message source.

<style>
  creatures quickly produce approve crevice nuclear moping
esoteric pernicious motion faith does embodies does
purify testament maximum exceeding centralism intellect prey
tidying welcomed traal impress tuneless athwart mansions
endures flames echo motion rooms alcohol rituals
etc.. etc.. etc..
</style>

That is followed usually by common image spam with a link "Click for
more" or "Go to details".

Anyone have a solution for this?  Can SA be trained to ignore whats in
between <style> tags?  What would that break?

Thanks,

- N