OpenAI has incentive to 'accidentally' allow toxic content through, so when they make the case that all models should be censored and make it safe, they can pull up the ladder behind them.