We present the first general protocol for secure multiparty computation which is scalable, in the sense that the amortized work per player does not grow, and in some natural settings even vanishes, with the number of players. Our protocol is secure against an active adversary which may adaptively corrupt up to some constant fraction of the players. The protocol can be implemented in a constant number rounds assuming the existence of a "computationally simple" pseudorandom generator, or in a small non-constant number of rounds assuming an arbitrary pseudorandom generator.