It's definitely a well-intentioned idea and a step in the right direction. But the detractors absolutely have a point about this being unenforceable.
Currently, you essentially need to have a PhD in computer science with a focus on machine learning to be able to take apart an algorithm like this to the point where its decisions can be effectively audited. And that would still be an incredibly time intensive and laborious task.
I think it would be much better to regulate which data can be used for training and then treat it's decisions like that of a hiring manager, auditing it's choices manually to ensure that diversity is not being indirectly filtered out.