AbstractTritium recovery by cold trapping in an intermediate NaK loop is considered for a self-cooled Pb-17Li blanket. The transient build-up of tritium inventory in the Pb-17Li and NaK loops is calculated for a test-plug to be used in NET. A steady-state tritium inventory of about two gramms is assessed for the blanket system including the cold traps.Hydrogen separation and release experiments in cold traps are going to start. Calculations of the temperature, velocity and concentration distributions in cold traps point to the influence of natural convection in mesh packed cold traps. First hydrogen release experiments were performed with hydride powder and hydride crystals. Different release mechanisms were observed during recovery from NaK which suggest that tritium recovery from drained cold traps should be preferred.