In commercial transactions, an intermediary might be involved to help transacting parties to conduct their business. Nevertheless, the intermediary may not be fully trusted. In this paper, we introduce the concept of intermediary (or agent) in a non-repudiation protocol, define the aims of intermediary non-repudiation protocols, and analyze their security requirements. We present a simple scenario with only one recipient, followed by a more complicated framework where multiple recipients are involved and collusion between them is possible.