It's absolutely believable that at first he thought the best way to safeguard AI was to get rid of the main advocate for profit-seeking at OpenAI, then when that person "fell upward" into a position where he'd have fewer constraints, to regret that decision.