Relishing new Fickling features for securing ML systems

Oh this is going to get interesting, grab the popcorn.

  OpenAI: "Our stuff is safe. So long as normal, good people only use
          it for normal, good things."

  NYTimes: "But what about bad people using it for bad things?"

  OpenAI: "That could never happen with our 'safeguards' ".

  NYTimes: "Challenge accepted." hack, hack, hack....

  OpenAI: "Oh noes! Our precious stuff!! You cheated, Filthy, evil
           hackers."