Amid concerns about biases, programmers have worked to make algorithms more benevolent. But while people have grown more trustful of benevolent AI, they’re unwilling to cooperate with it as they would with other people. This poses challenges for the future of AI-human relations.