Orientations of active antithetic faults can provide useful constraints on in situ strength of the seismogenic crust. We use LINSCAN, a new unsupervised learning algorithm for identifying quasi‐linear clusters of earthquakes, to map small‐scale strike‐slip faults in the Anza‐Borrego shear zone, Southern California. We identify 332 right‐ and left‐lateral faults having lengths between 0.1 and 3 km. The dihedral angles between all possible pairs of conjugate faults are nearly normally distributed around 70°, with a standard deviation of ∼30°. The observed dihedral angles are larger than those expected assuming optimal fault orientations and the coefficient of friction of 0.6–0.8, but similar to the distribution previously reported for the Ridgecrest area in the Eastern California Shear Zone. We show that the observed fault orientations can be explained by fault rotation away from the principal shortening axis due to a cumulated tectonic strain. Plain Language Summary: Small earthquakes can highlight the location and attitude of active faults at depth. We use a large set of earthquake locations and a novel algorithm to identify small faults, along with their orientations, and sense of slip. We find that faults with opposite sense of slip (the so‐called antithetic, or conjugate faults) are at nearly right angles to each other. For newly created faults, such a configuration would imply that friction is almost negligible. We suggest that the high‐angle conjugate faults instead result from fault rotation due to long‐term tectonic deformation. Key Points: We use a new algorithm to identify quasi‐linear clusters of micro‐earthquakes associated with active strike‐slip faults in a trans‐tensional region south of the Salton Sea, Southern CaliforniaThe observed dihedral angles between right‐ and left‐lateral faults show a broad distribution with a peak around 70°Non‐optimal fault orientations can be explained by tectonic rotation due to a long‐term slip on a more mature system of right‐lateral faults [ABSTRACT FROM AUTHOR]