My point was, ChatGPT and Large Language models are great at generating new fictional content, like a new TV show, but often wrong when faced with more factual questions, like "How to best handle a court case and provide a document to the court about the innocents of our client"
Just trying to get chatGPT to write simple code is amusing, if you know that what it spews is incorrect. In the hands of a non-programmer, relying on ChatGPT to generate an application would lead to folly.
If you are talking about their phones, they have to be kind of obviously recording, holding their phones up while these glasses may or may not be just passively recording.