Amnesty International denounces ‘AI’ breaches of human rights

CHECKING out bookmarks, the Freelance finds a report from Amnesty International that merits a proper look. Unlawful by design: Exposing the human rights costs of generative AI introduces itself as examining how “how standalone generative AI systems, based on unlawful web scraping, are in conflict with international human rights law (IHRL) and standards through their design, development and deployment”.

And it is not happy with them. We do recommend that you read all 44 pages. We nevertheless extract some key sentences on various fields of IHRL:

• The right to privacy: “Through... the use of web scrapers for collecting training data, some generative AI tools risk violating privacy laws and the rights enshrined within them. This may depend on whether the collection and processing of data used to build the tool has been conducted consensually or whether the individual and community right to privacy has been violated in the process, whether it serves legitimate interest, whether it is accurate and secure, and whether individuals may exercise their rights under data protection law.”
• The right to equality and non-discrimination: “Just as AI-driven surveillance tools such as facial recognition have been subject to calls for prohibition due to the high risk of profiling according to ethnicity, race, national origin, gender identity and other prohibited grounds, which is often the basis for unlawful discrimination, generative AI tools that rely on massive data collection exercises present similar risks of discrimination.”
• The right to freedom of expression: “Private actors also have a responsibility to respect human rights, as established under international human rights standards such as the UN Guiding Principles. Content moderation on the internet, including social media platforms, can be considered one such measure by which private actors can fulfil their responsibility to respect human rights in the context of free expression. Content moderation must be conducted in a manner ensuring respect for the right to freedom of expression, while addressing rampant advocacy of hatred – a systemic issue on social media platforms.”
• Freedom of thought: “A 2024 report published by the Centre for International Governance Innovation draws attention to the right to freedom of thought as an increasingly salient right to be considered against the backdrop not only of emerging brain-computer interfaces, but also data-driven persuasive technologies, eye-tracking and pupillometry (measuring the movement and reaction of the eye), scene understanding (ability to comprehend and analyse a visual scene), and the predictive power of AI language models. While the right to freedom of thought is an oft-neglected right in international law, it... is an increasingly salient and important tool in assessing the rapid adoption of generative AI technologies in particular.”
• Business and human rights standards: “Transparency is a key component of human rights due diligence. As the UN Guiding Principles [on Business and Human Rights] make clear, companies “need to know and show that they respect human rights.”

The report’s first two top-level recommendations to states align very closely with our interests:

• “Implement a prohibition on standalone generative AI systems that have been built using unlawful web scraping, defined as the bulk and mass collection of training data through the World Wide Web, without protection against non-consensual collection of personal data”; and
• “Enact legislation requiring transparency regarding training data collection practices and accountability across AI supply chains...”

Amnesty International’s recommendations to companies open with “Cease the practice of unlawful non-consensual web scraping of personal data for AI training purposes” and continue with seven more ways in which they should respect human rights law.

© Mike Holderness

Unlawful by design amnesty.org

ʻNo means no’: Freelances push back on ʻAI’ the long read on our 8 June Parliament meeting

'Artificial Intelligence' our coverage to date