How should we rethink the take-home assessment?

ChatGPT is an awesome tool for developers - it’s only a matter of time before it gets mass adoption. However, I think its causing a bit of a shake-up in how interviewing engineers is done.

I'm particularly interested in understanding its impact on take-home assessments. Personally, I'm a fan of take-homes (as long as they are done correctly). One of the main benefits of take-home assignments is that they don't create the same anxiety as live interviews. Yet, I'm starting to see companies react to ChatGPT by tightening the proctoring constraints of their assessments - completing an assignment in a limited timeframe, using a specific online IDE, disabling copy-paste functions, and so on. But, this kind of approach seems to defeat the purpose of take-home tasks.

How should we modify take-home assessments due to ChatGPT, without resorting to stringent proctoring?