“The outputs of automated systems may become biased if they arise from biased data input or are trained on biased data, or bias may arise due to a biased originating source with no human corrective
action taken,” the ANA’s guidelines state. “Having a diverse development team can help protect against this potential bias.”