While I haven't taken the time to research further, I figured I'd try to brick the algorithm by speaking absolute gibberish. I also tell everyone else I know. Since machine learning is still "dumb", I figured that enough people babbling incoherently would require the spammers to hand-scrub the data or make more training rules, both requiring more work.

No idea if it worked, but it was the least I could do to halt progress in the wrong direction.