Is It a Simplification of AI?

Some organizations are trying to pursue that, well, why don't we develop AIs that can help us develop better AI? And I wondered what you thought about the approach of trying to use it as a tool to help us design a better version of a future. So open AI is using that approach, right? But it feels like it's a bit of a catch-22. It's AIsafety Complete Problem. You cannot use a malevolent AI to help you create a safe AI at least not in any verifiable way. As depressing as it might be, that sounds like that might be the outcome.