GPT-4 emergent behavior, release not endorsed by red team

The prospect of an autonomic LLM model making money by whatever means possible in order to run more copies of itself could be a bit of a problem. I wonder how long until the first 100% AI ransomware attack happens?