I’ve been working with a company that provides inference as a service to customers across various industries. One thing that’s struck me is the sheer amount of information banks request through model governance. I’m talking about specific features used in the model, data sources, detailed explanations of our methodology, and even the time frame for train/test/val sets down to the day.
I’m left wondering if my privacy team is being overly cautious or if this is just the new normal. The back-and-forth with customers is a huge time suck, and I’m not sure if we’re doing things right.
So, I turned to the community for advice. Is this level of scrutiny normal, or do we need to improve our reporting to customers who are concerned about these kinds of things?
From what I’ve gathered, banks want to know everything about our models, from the features used to the metrics for evaluating performance. It’s like they want to recreate the model themselves. I get it – they’re risk-averse and need to ensure their systems are secure. But is this level of transparency really necessary?
I’d love to hear from others in the field: how do you handle model governance requests? Are there any resources out there that show what’s industry standard? How do you balance transparency with protecting your intellectual property?
Let’s get the conversation started!