Mr. John Sherwin:

That is a difficult question to answer. The testing of these systems before they are released and deployed needs to be significantly more effective than it is at the moment. Systems are being released with bias and are based on casual testing by public testers, who may be keen to expose those problems and post them on social media in order that they become apparent very quickly. To have chatbots, GhatGPT, Grok or whatever, where they are being used for advice is one thing, but when there is an AI-powered service that is going to determine care decisions, that needs a significantly higher level of testing. There is a testing component to it that is really important before it gets released, and also a commitment from the company for ongoing assessment when it is released.