AI’s human rights impacts debated at Oxford summit

Human rights could provide a common framework

The development and use of AI systems, along with the resources required to operate them, have a significant impact on people and the environment, and in the future on outer space as well. Researchers at the Oxford Institute for Ethics in AI said a human rights-based approach could provide a common language for addressing these issues. It can help identify needs and interests, formulate specific demands and hold those responsible to account.

Discussions at the AI and Human Rights Summit, held in Oxford on October 5, 2026, were organised around three questions: what is missing, what needs to change and how can change be achieved? The organisers, Dr Caroline Emmer De Albuquerque Green, Research Director at the Institute for Ethics in AI, and Professor Yuval Shany, Lead Fellow of the Accelerator Fellowship Programme, summarised the points raised by participants.

According to the report, technology companies now play a decisive role in shaping the pace and direction of development. The rapid spread of AI and its social effects are making it difficult for traditional human rights protection mechanisms to adapt to new economic, political and social power dynamics.

Calls grow for binding regulation and independent evidence

Several participants argued that the legal challenges posed by AI cannot be addressed through non-binding standards alone. The world should gradually move towards a comprehensive global treaty on AI and human rights. Since harm caused or facilitated by AI often crosses national borders, mutual legal assistance agreements may also be needed.

Having rules in place is not enough. Their enforcement is also crucial. According to the report, weak enforcement typically harms people with little power, while political and corporate leaders can often operate without consequences. Questions were also raised about whether existing laws can evolve quickly enough, for example to address the relationship between AI-driven digital manipulation and freedom of thought.

The summit also identified the lack of reliable, independent evidence as a significant problem. Participants said comparable data was needed on how AI affects people’s lives, particularly children and vulnerable groups. Research should examine both beneficial and harmful effects, as well as the climate, energy, water and land-use consequences of data centres. One proposal was to require AI developers and users to support independent research and publish its findings.

Responsibility should move up the AI chain

Some participants argued that responsibility should be shifted to the people and organisations that determine the objectives and parameters of AI systems. This would differ from an approach that places primary responsibility on the so-called human-in-the-loop who monitors the system’s output. Several also suggested that personal civil and criminal liability rules may be needed to cover both intentional and unintentional harm.

Transparency was a recurring topic during the discussion, as were Meta’s child safety cases, which participants said served as a reminder that major technology companies have repeatedly resisted transparency and accountability. The report highlights the management of private power as one of its central questions.

Environmental protection also received particular attention. The discussion covered the rights of people living near data centres, the effects of energy demand on local infrastructure and the risk of reduced access to clean drinking water. Participants said simply relocating data centres from the global north to the global south would shift environmental burdens onto other communities and energy grids.

According to the organisers, children often appear only at the margins of the debate, even though they will live with the effects of AI on their lives for the longest time. Participants therefore called for better oversight, clear human control and approaches designed to protect the full range of human cognitive abilities.