Leiden Declaration: AI is challenging the core values of mathematics(universiteitleiden.nl)
universiteitleiden.nl
Leiden Declaration: AI is challenging the core values of mathematics
https://www.universiteitleiden.nl/en/news/2026/06/leiden-declaration-warns-ai-is-challenging-the-core-values-of-mathematics
1. Unreliable results
True for human generated results too - how many "proofs" of Fermat's Last Theorem were proposed then discarded when a mistake was found, before Wiles?
2. Lack of proper attribution and violation of copyright
I have trouble understanding the difference between a human reading and digesting copyrighted materials (i.e. "training") and AI training. As long as either case doesn't reproduce this material word-for-word without attribution (which is why we have a word for this - plagiarism.) OK, maybe an electronic thinking tool makes plagiarism easier if you are careless. A power saw is more dangerous than a hand saw, too.
3. Dependence and inequality
Any fancy automation tool brings this risk. Is a Matlab subscription or access to a data center "inequitable"? And yeah, if you live in a place where you need a car, then you have a mobility dependence. You can still walk, but it isn't practical.
4. Overhyping of results
I don't see a difference between human or AI generated here, although the newness of AI does lead to people being more interested at the moment. The shine will wear off eventually.
5. Loss of autonomy
I just don't understand why AI is a problem here - the "autonomy" here is about who sets what topics are interesting in mathematics. Is it because every amateur now has a new powerful tool that once only existed in the brains of a few adepts? So they (or AI itself) might have input into what problems are important?
So overall I think this isn't a very effective list of risks. Everyone is nervous that the unique thinking abilities that set mathematicians apart (and I completely get this as a physicist) are being eroded. Until the reliability of the output of AI systems improves by several nines, we need humans in the loop to make sure the output is correct (Point 1.) But given that AI is improving rapidly, extrapolation to the day where humans are effectively out of business is now something that needs to be considered. Just declaring we need humans to do mathematics research isn't going to be enough if that day ever arrives.
Edit: formatting