Secure audio teleconferencing is a multi-point co-&cation service which uses enuyption to prevent eavesdroppers from listening to the speech signals. Its greatest vulnerability is the audio bridgethat component which combines the conferees' speech signals and returns the result to them. A new secure teleconferencing system is proposed here. It fits the public telephone network by eliminating the need for the conferees t o share their secrete with the bridge. It combines a simplified ('instantaneous') bridging technique with secure bridging ideas previously suggested in the literature, overcoming thek main practical disadvantages. In particular, it is not restricting the audio signals to be coded by linear PCM, a technique which is wasteful in terms of bit-rate. Rather, it enables the use of conventional p-law and A-law PCM, well as vector quantized PCM, thus can be used with a conventional 64kb/s digital channel.