The Carlson Lab at the Yale School of Public Health has done some research and held some discussions, and posted The only reason you’ll ever need not to write with AI.

Our lab has agreed on a blanket ban on the use of generative AI for writing….Using generative AI inherently exposes you to the risk of career-ending accusations of plagiarism.

Even though there’s a lot of pressure on academics to adopt AI or be left behind!!1! this is a lot bigger deal. Especially for this particular research team.

They’re a public health lab, which puts them into an unavoidable adversarial relationship with the anti-public-health oligarchs who run the “AI” companies. The “E” in TESCREAL doesn’t stand for “eugenics,” but the political movement behind “AI” is more likely to believe in improving society by weeding out the “unfit” recipients of health misinformation than in public health as we know it.

More from the blog post:

Plagiarism accusations are now a core part of the right-wing harassment toolkit, and knowing that those accusations were made in bad faith or for political reasons does not make them disappear.

It’s safest to assume that a future well-connected accuser will be able to get access to large quantities of text, for practical purposes all of it, and the computing resources to look for a pattern match between anything that the targeted person has ever written and anything that anyone else has ever written. They’ll be able to do the search much better than the targeted person could, even using the best commercial “plagiarism detection” software and services.

Related: On the non-use of AI in my writing process by Charles Stross.

If I was crazy enough to feed the outline of a story I was working on as a prompt to ChatGPT or Claude in hope of getting the stochastic parrot to do my homework for me, then it would be only my own fault and nobody else’s if the next model from the company in question was trained on my book outline and could reproduce part or all of it for someone else.

Authors of scientific papers don’t have the choice to keep their work out of training sets in most cases. As soon as it’s published or on a preprint server it will get crawled, either with the publisher’s cooperation or not.