Due to the texture formed in cold/hot rolled forming process, anisotropy is a key issue not only in modeling of plastic deformation but also in characterization of fracture behavior. In this study, an anisotropic ductile fracture criterion is developed by introducing anisotropic parameters into the weight function of an uncoupled shear ductile fracture criterion. The proposed anisotropic ductile fracture is applied to describe the anisotropic characteristics in the ductile fracture of AA6082-T6. Ductile fracture behavior of AA6082 is experimentally investigated at the different loading conditions: shear by in-plane torsion test, uniaxial tension by specimens with a central hole, plane strain tension by notched specimens, and the balanced biaxial tension by the Nakajima test. In-plane torsion and tension tests with a central hole and notch are conducted along three directions: rolling direction, diagonal direction and transverse direction. Specimen deformations during the tests are recorded and fracture strains are measured by digital image correlation approach. The measured fracture strains are then utilized to calibrate the parameters in the proposed ductile fracture criterion. With the calibrated ductile fracture criterion, the fracture locus and anisotropic ductile fracture in various loading conditions are predicted and compared with experimental measurement and those predicted by linearly transformed anisotropic fracture model to investigate the predictability of the proposed ductile fracture criterion. The comparison demonstrates that the anisotropic fracture of AA6082 is predicted by the proposed criterion with good agreement in the different loading directions of shear, uniaxial tension, plane strain tension, and the balanced biaxial tension. Considering the high accuracy of the proposed ductile fracture criterion, it is expected that the proposed anisotropic ductile fracture criterion can improve the reliability of failure prediction in metal forming for materials with strong directionality in fracture. (C) 2019 Elsevier Ltd. All rights reserved.