Why wouldn't it occur to OpenAI to run their models to secure their own systems? Seems like a clown show.