What could go wrong, given that OpenAI is currently under a court order requiring retention of (all? unsure how applies to new products) chatGPT transcripts and their possible disclosure to plaintiffs?