Dataset Discrimination In Government Surveillance: A Threat To Equality And Justice

The development of complex algorithms enabled the creation of artificial intelligence (AI) models that can act rationally and encode thousands of variables across millions of data points. These models have proven useful, including for the public sector. For example, machine learning has been used to predict European Court of Human Rights decisions; in Brazil, the Victor project is an Al system designed to identify general repercussions of pending cases in the Brazilian Federal Supreme Court to enhance analysis efficiency. These tools have also been applied in government surveillance to identify individuals flagged as suspicious or dangerous, monitor crowds for potential threats, and help security staff locate lost children. The dataset discrimination debate is aligned with G20 priorities, as shown by the Sherpa Track’s Digital Economy Working Group agenda. Also, some of its members can contribute with their own experience, as the European Union’s path about the Al Act. The potential of Al-powered facial recognition in public spaces to enhance security and law enforcement is undeniable. Still, there are critical concerns about Al-based facial recognition in public spaces: discrimination and bias, lack of transparency and accountability, privacy violations, data protection, and cybersecurity. The G20 should establish a commission to investigate the ethical implications of Al and its associated machine learning and deep learning technologies, develop a model framework for using facial recognition in public spaces, outline common principles and minimum standards to guide national legislation, and launch a data governance initiative to promote harmonizing data protection standards.