We present the scientific motivation for future space tests of the equivalence principle, and in particular the universality of free fall, at the $10^{-17}$ level or better. Two possible mission scenarios, one based on quantum technologies, the other on electrostatic accelerometers, that could reach that goal are briefly discussed.

Over five years we have compared the hyperfine frequencies of 133Cs and 87Rb atoms in their electronic ground state using several laser cooled 133Cs and 87Rb atomic fountains with an accuracy of ~10^{-15}. These measurements set a stringent upper bound to a possible fractional time variation of the ratio between the two frequencies : (d/dt)ln(nu_Rb/nu_Cs)=(0.2 +/- 7.0)*10^{-16} yr^{-1} (1 sigma uncertainty). The same limit applies to a possible variation of the quantity (mu_Rb/mu_Cs)*alpha^{-0.44}, which involves the ratio of nuclear magnetic moments and the fine structure constant.