Landlords, employers, courts, agencies, insurance companies, police, social media platforms, and others use AI systems to make automated decisions that may impact people’s rights. These systems often reflect embedded bias. Limiting the risk of harm begins, but doesn’t end, with meaningful transparency, due process for people subjected to automated decision-making (ADM), and impact assessments regarding potential bias and other harms.  Government use of ADM in particular must meet a very high bar: systems used to support consequential decisions must be interpretable, explainable and auditable. ADM should respect international human rights standards. Many of the most popular AI architectures in use today do not meet those goals. 

EFF fights automated oppression alongside community members, lawmakers, and litigators, from San Francisco to Strasbourg.