By Mark Roth-Whitworth: I mentioned on File 770 last week that I’d just signed a contract for two novels. One of the things that I was asking for was to have them insert on the disclaimer page (you know, “This is a work of fiction…”) a statement like “This work may not be used to train generative AI” (chatbots, that is). Their lawyer said no, and after thinking about it, I decided that this needed a much broader solution. So I just sent this email to my Rep and both Senators, and it’s my part of the answer to chatbots and copyrights.

Hon. [Rep.|Sen]

I was negotiating a literary contract for two novels with a publisher last week, and the issue of AI training came up. I had been asking for them to print, on the disclaimer page, that the work could not be used for such training. Their lawyer tells them that they could not police large companies training generative AIs – chatbots – and so only put it in the contract that they would not.

The training of chatbots is a huge issue for artists, writers, and others, as much is done by simply scraping the Web, ignoring copyright notices. I believe that we can begin to address this issue by what I am requesting: you introduce as a bill to modify existing copyright law. My draft, and I am certainly not a lawyer, would be the following:

“Works in copyright may not be used to train generative AIs without explicit permission from the copyright owner.”

The impact of this would be that the corporations running and training them would be required to search whatever data they are using for copyright information, before they use it for the training. Further, should there be any question of copyright violation in this manner, if the chatbot is capable of printing out explicit copies of the work in question, it would be a prima facie case of violation. That would make it possible for creators to sue without being millionaires.

Thank you for your attention. I look forward to your response.

mark