Data pipelines powering generative AI systems rooted in invasions of privacy (www.amnesty.org)

Amnesty International's recent briefing, "Unlawful by Design: Exposing the Human Rights Costs of Generative AI," reveals significant concerns regarding the unethical data practices underpinning generative AI systems. The report claims that companies utilize illegal web scraping to gather vast amounts of personal data—such as images and social media activity—without consent, raising serious privacy violations and human rights issues. Key generative AI models, including those from OpenAI, Google, and Meta, have been implicated in perpetuating biases and exclusionary outputs due to the polluted datasets that often mirror real-world prejudices, especially against marginalized communities. The implications extend beyond privacy violations, as the increasing infrastructure demands of these technologies are amplifying environmental costs, evidenced by a rapid rise in greenhouse gas emissions linked to data centers. The call to action from Amnesty emphasizes the necessity for regulatory frameworks that prohibit the use of unlawfully sourced training data and hold corporations accountable for their human rights impacts. The briefing underscores the urgent need for a reevaluation of the technological development path taken by these companies to align with ethical standards and human rights considerations.