Googling reveals these lectures slides of Ken Ribet, which say the following.

After Serre's article on elliptic curves was written in the early 1970s, his techniques were generalized and extended in different directions. In particular, Serre and Swinnerton-Dyer began to study the mod $\ell$ representations of $\text{Gal}(\overline{\textbf{Q}}/\textbf{Q})$ attached to the cusp form $\Delta$ of weight $12$ on $\textbf{SL}(2, \textbf{Z})$.

These representations weren't always with us: It was only in his 1967-1968 DPP seminar on modular forms ("Une interprétation des congruences relatives à la fonction $\tau$ de Ramanujan") that Serre proposed the possibility of linking Galois representations to holomorphic modular forms that are eigenforms for Hecke operators. Almost immediately afterwards, P. Deligne constructed the representations whose existence was conjectured by Serre.

This suggests the specific report is the following, see here.

Serre, Jean-Pierre. "Une interprétation des congruences relatives à la fonction $\tau$ de Ramanujan." *Séminaire Delange-Pisot-Poitou*. *Théorie des nombres* 9.1 (1967-1968): 1-17.

A pithy one sentence review of the report by B. Stolt on MathSciNet says the following.

The author makes a survey of results on Ramanujan's $\tau$-function.