A new direct adaptive algorithm for feedforward active noise control (ANC) is proposed in a general case when the primary and secondary path responses are all uncertain. In order to reduce the actual canceling error, the two artificial errors are introduced and are forced into zero by adjusting three adaptive FIR filters in an on-line manner. The distinctive feature of the method is that the adjusted parameters do not have to converge to the true FIR parameters but some constants, then the reduction of the two errors can attain the canceling at the objective point. The effectiveness of the proposed algorithm is validated in experiments using an air duct.