After the release of the present dosimetry system DS86 in 1987, measurements have shown that DS86 may substantially underestimate thermal neutron fluences at large distances (>1,000 m) from the hypocenter in Hiroshima. This discrepancy casts doubts on the DS86 neutron source term and, consequently, the survivors' estimated neutron doses. However, the doses were caused mainly by fast neutrons. To determine retrospectively fast neutron fluences in Hiroshima, the reaction 63Cu(n, p)63Ni can be used, if adequate copper samples can be found. Measuring 63Ni (half life 100 y) in Hiroshima samples requires a very sensitive technique, such as accelerator mass spectrometry (AMS), because of the relatively small amounts of 63Ni expected (approximately 10(5)-10(6) atoms per gram of copper). Experiments performed at Lawrence Livermore National Laboratory have demonstrated in 1996 that AMS can be used to measure 63Ni in Hiroshima copper samples. Subsequently, a collaboration was established with the Technical University of Munich in view of its potential to perform more sensitive measurements of 63Ni than the Livermore facility and in the interest of interlaboratory validation. This paper presents the progress made at the Munich facility in the measurement of 63Ni by AMS. The Munich accelerator mass spectrometry facility is a combination of a high energy tandem accelerator and a detection system featuring a gas-filled magnet. It is designed for high sensitivity measurements of long-lived radioisotopes. Optimization of the ion source setup has further improved the sensitivity for 63Ni by reducing the background level of the 63Cu isobar interference by about two orders of magnitude. Current background levels correspond to a ratio of 63Ni/Ni<2x10(-14) and suggest that, with adequate copper samples, the assessment of fast neutron fluences in Hiroshima and Nagasaki is possible for ground distances of up to 1500 m, and--under favorable conditions--even beyond. To demonstrate this capability, we have measured successfully 6Ni/Ni ratios as low as (3.5 +/- 0.6) x 10(-13). The latter are, based on DS86, representative of a ratio expected from a typical Hiroshima copper sample at about 1,300-m ground range.