In this paper, I use a framework from computational linguistics, the Rational Speech Act framework, to model deceptive probabilistic communication. This account allows agents to discount for the biases they perceive their interlocutors to have. This way, agents can update their credences with the perceived interests of others in mind.