Amnesty International has issued a stark warning about the human rights risks tied to generative AI systems, stating that they are “unlawful by design.” The organization’s latest briefing reveals that companies are scraping vast amounts of online data — including personal information, images, and social media activity — without consent to train models such as OpenAI’s GPT‑3, Google’s Gemini, Meta’s Llama, and others.
The report argues that this practice amounts to mass privacy violations and entrenches racial, gender, and cultural biases. By relying on non‑consensual data extraction, generative AI systems amplify harmful stereotypes and discriminatory outputs, disproportionately affecting historically marginalized communities.
Beyond privacy concerns, Amnesty highlights the environmental costs of AI production. Larger models demand energy‑intensive chips and massive data centers, driving up emissions and water usage. Communities in Chile, Mexico, and the United States have resisted data center projects due to droughts and electricity shortages. Reports from Google and Microsoft show significant increases in greenhouse gas emissions linked to AI infrastructure.Amnesty International is calling on governments to prohibit generative AI systems built on unlawful web scraping and to hold companies accountable for human rights abuses tied to their design. The organization insists that a different trajectory for technology development is possible if authorities act urgently to enforce privacy, ethics, and sustainability standards.
