In this study we examine the psychometric characteristics and the predictive validity of the AISS-17, the French version of the Arnett Inventory of Sensation Seeking. Two samples were used, one of which enabled us to carry out prospective monitoring over a period of six months. Our study provides data that support the validity of the AISS-17. This inventory includes the same two factors as the AISS (Novelty and Intensity). It has good concurrent and convergent validity, and its nomological network is comparable to that of the AISS and to the SSS of Zuckerman. However, our study confirms the low reliability of the AISS. We also showed that the AISS-17 has good predictive validity. We observed a significant cross-lagged effect of the AISS-17 at T1 on risk taking at T2, when controlled for age, gender and risk-taking at T1.