Author: Aaina Agarwal, Patrick Hall, Sara Jordan, and Brenda Leong
Publisher: Future of Privacy Forum
Publication Year: 2021
Summary: The following article focuses on looking at the ethics of AI use from a lawyer’s perspective, providing some of the potential legal issues that could arise with AI misuse. The article brings up 5 different aspects of AI to be aware of. Firstly, machine learning systems are probabilistic. No machine learning model can make predictions accurately on every single occasion, so it is essential to mitigate some of that risk by ensuring that your training data is as robust as possible. The Federal Reserve has a model risk management guidance document which is a great resource to ensure that as much risk as possible is being accounted for during model creation and deployment. The resource then discusses AI transparency, specifically ensuring that your customers understand what a particular model is doing, and model bias, how it can often manifest in unexpected ways. The article ends with a forward-looking perspective, discussing how we are at the early stages of professional standards and law for AI. However, there has been a lot of progress in this area, including the creation of statistical tests to detect algorithmic discrimination.