Major generative AI systems are continuing to get away with violating human rights

Report cover titled Unlawful by Design: Exposing the Human Rights Costs of Generative AI, with the main title in white text on a black bar and the subtitle in black text on white bars. Against a bright yellow background, a large blue-toned robotic hand reaches down from the upper left, with thin pink and black strings descending from it like a puppeteer's controls. At the bottom, the strings connect to a row of black-and-white photos of workers wearing headsets at computer workstations in cubicles, each marked with a pink dot, suggesting the human labor controlled by or behind the AI system.

Amnesty International published a report on how major generative AI (genAI) systems severely violate privacy laws, further opening up the risk of mass human rights abuse.

Amnesty found that privacy infringement was a deliberate design choice in the training of genAI models. As these systems are trained on illegally scraped data from the internet, this means any social media post, online blog contribution, or comment under a video is used without the creator/user’s consent or knowledge.

In addition to the environmental exploitation needed to train such large-scale models, Amnesty highlighted how genAI outputs perpetuate discrimination and existing inequality. A large portion of internet data is deeply discriminatory, especially along racial and gendered lines.

Amnesty reached out to tech companies (Intel, VMware, Google, OpenAI, Meta, Stability AI, Midjourney, and DeepSeek) for a statement. Those that responded highlighted various future environmental ‘water positive’ goals, and the ‘importance of human rights’, but none delved into how privacy infringement and discriminatory output would be avoided. OpenAI went so far as to cite its ‘efforts’ to protect rights holders, such as vulnerable groups, in the context of covert influence activity by state authorities—claims which Amnesty was not able to verify as legitimate.

This comes as no surprise, given that big tech companies lack any moral or socially responsible backbone. We know this from their contracts with Palantir, the company abetting genocide through AI-weapons systems.

Human rights will continue to be violated if no significant changes are made to the deliberate design choices. Amnesty calls for clear frameworks for accountability and proper justice for those wronged, and an internationally mandated prohibition of these systems – whether proprietary or open source – if trained with illegally scraped data.

You can download the full report here (PDF).

See: Unlawful by design: Exposing the human rights costs of generative AI at Amnesty.

Image from the cover of Amnesty’s report.

Comments are closed.

Proudly powered by WordPress | Theme: Baskerville 2 by Anders Noren.

Up ↑