Rights group raises concerns about unlawful data collection systems to train generative AI

Amnesty Worldwide reported on Thursday that tech corporations used illegal internet scraping to gather giant volumes of on-line information for the event of generative synthetic intelligence (AI) fashions, in violation of the precise to privateness and different human rights requirements. The group known as for the prohibition of such information assortment methods and urged governments to intervene to control these practices.

In a report documenting the dangers related to the large-scale information assortment and processing methods deployed by corporations equivalent to Google, Meta, and OpenAI, Amnesty Worldwide said that standalone generative AI methods depend on illegal internet scraping to extract in depth quantities of customers’ on-line information to coach their generative AI fashions, which makes these methods illegal by design and deployment.

Based on Likhita Banerji, Head of the Algorithmic Accountability Lab at Amnesty Worldwide, these information scraping methods extracted data from billions of public on-line posts worldwide with out the specific consent of internet customers, violating their proper to privateness enshrined in elementary human rights treaties and particular resolutions on the precise to privateness within the digital age. The group additionally mentioned that the generative AI methods perpetuated racial and gender biases in addition to discrimination, which is a mirrored image of real-world biases and cultural stereotypes inherent to the coaching information pulled from the net.

Moreover, Amnesty Worldwide highlighted that the fast improvement of generative AI fashions had critical environmental penalties. Firstly, the event of generative AI requires constructing information centres to accommodate AI servers, which eat substantial water assets throughout each building and operation, whereas water is more and more scarce in lots of locations. These information centres additionally produce digital waste, which frequently accommodates hazardous substances equivalent to mercury and lead. Secondly, AI-related infrastructure closely depends on essential minerals and uncommon components, which are sometimes mined unsustainably.

Consequently, Amnesty Worldwide known as on tech corporations to stop the illegal mass assortment of knowledge to coach their standalone generative AI fashions and urged states to carry corporations accountable for human rights abuses associated to AI instruments design and enterprise decisions.

There was a debate concerning the challenges associated to AI improvement. Key issues included the misuse of AI in surveillance applied sciences, nonconsensual information extraction, information leakage, inferential profiling of individuals, and the era of artificial media able to influencing particular person behaviour, significantly amongst weak teams equivalent to kids, which prompted nationwide and worldwide efforts to implement applicable laws.

In response, a number of international locations, specifically Brazil and Vietnam, have enacted particular laws to deal with the difficulty of security in AI use. In February 2026, the UN Secretary Basic emphasised the necessity to preserve AI governance and regulation with out impeding innovation, given the growing use of AI applied sciences by the army to commit human rights violations.