Is there another controversial statement by Grothendieck apart from 57 being prime? There is a well-known story about Grothendieck being asked to explain concretely some result involving prime numbers and of his answering  "You mean an actual number? All right, take 57".
See here.
Unfortunately there seems to be no written trace of this anecdote and it it is not clear whether it happened or not.
But did he make a written assertion in the same vein?
 A: Yes, he did.
In his  Récoltes et semailles, volume I posthumously edited by Gallimard in 2022 he reminisces (in footnote 68, page 522) about his first encounter with $\pi$ as a child:
"La valeur approchée 344/133 trouvée dans un livre (...) m'avait frappé-elle était si jolie que j'avais du mal à croire qu'elle ne soit qu'approchée!"
which I translate as:
"The approximate value 344/133 found in a book (...) had struck me-it was so pretty that I could hardly believe that it was only approximate !"
This is interesting because $344/133=2.586466...$, certainly the worst approximation of $\pi$ in the history of Mathematics!
Needless to to say his unnamed book certainly had $\pi\cong355/113=3.14159...$, with all digits correct.
What I find amusing is that neither Grothendieck in his fifties when he wrote Récoltes et semailles, nor the people at Gallimard, nor the numerous mathematicians who retyped the manuscript, nor any of my colleagues who read the book noticed this egregious blunder...
As a bonus, here is the retyped manuscript , with the footnote on page 318.
