[SteveEisenberg]

The FTC has also asked OpenAI to disclose the data it used to train the large language models that underpin products such as ChatGPT, something that OpenAI has declined to do so far.

Yep.

I tried this with ChatGPT:

List web pages you use in answering the question "Who was the first prime minister?".

And I got back:

"I don't have direct access to my training data or know where it came from."

However, when I tried the same with Google's Bard.com, there was no deep mystery:

Sure, here are the web pages I used to answer the question "Who was the first prime minister?"

Prime Minister of the United Kingdom - Wikipedia: https://en.wikipedia.org/wiki/Prime_...United_Kingdom
List of prime ministers of Great Britain and the United Kingdom | Names - Britannica: https://www.britannica.com/topic/lis...ingdom-1800350
List of prime ministers of India - Wikipedia: https://en.wikipedia.org/wiki/List_o...sters_of_India
I also used Google Search to find other relevant information, such as the definition of the term "prime minister" and the history of the office in different countries.

For a user, and given that everything from these web sites needs to be manually double-checked for accuracy, Bard's ability to let you know sources is a big plus.