The results of y-ray observations by the COS-B satellite lend support to the suggestion that a class of y-ray sources comprises extended sources, associated with selected active regions in the Galaxy, and in which supernova remnants and/or strong winds from young massive stars are at work. In this framework, y-rays come either from bremsstrahlung radiation from relativistic electrons, or from n0 decay from in situ accelerated relativistic protons. Specifically, a case is made for the identification of the Carina complex with the y-ray source 2CG 288-00. By using a simplified model for the Carina complex, it is shown that, in this source, y-rays may be the result of proton acceleration by stellar winds, followed by confinement by resonant Alfven-wave scattering in the giant Hu region NGC 3372. The confinement is efficient enough for in situ accelerated cosmic rays to dominate ambient cosmic rays. The model predicts that the source should be identified with NGC 3372, with little contribution from the adjacent molecular cloud. The required proton acceleration efficiency is of the order of 10% . In the Orion complex, the same model leads to the conclusion that the confinement is inefficient, and that, as a consequence, ambient cosmic rays dominate in situ accelerated cosmic rays, as observed.