Even if there is an accepted answer, I would like to question the hypothesis.
According to
- Ralf Krömer, La « machine de Grothendieck » se fonde-t-elle seulement sur des vocables métamathématiques? Bourbaki et les catégories au cours des années cinquante, Revue d'histoire des mathématiques, Volume 12 (2006) no. 1, pp. 119-162. publisher, HAL, Numdam,
universes (in categorical foundations) were introduced as a Bourbaki "rédaction" ("Sur la formalisation des Catégories et foncteurs" no 307) by Grothendieck in 1958(?) as an alternative of the proposed foundations by (Formalisation des classes et catégories) D.Lacombe inside the group.
It appears that the archives Bourbaki does not have an electronic copy of it. But the cited reference has the appropriate excerpt in pg 149.