Summit calls for human rights safeguards on AI

Human rights experts called for stronger legal protections against the risks posed by artificial intelligence (AI) at an Oxford AI and Human Rights Summit, with speakers arguing that human rights standards must be embedded into AI governance and technology. 

The summit was hosted at the newly redeveloped Rhodes House in Oxford and organised by the Institute for Ethics in AI. It built upon Professor Yuval Shany’s consultation on a white paper, published in 2025, on the need for and feasibility of an international AI Bill of Human Rights, which proposes ways to adapt existing human rights standards so as to better address some of the unique challenges posed by AI systems and their uses. 

An opening video address from the United Nations High Commissioner for Human Rights Volker Türk outlined how AI policy can keep pace with a fast-developing technology. He said that “great power demands responsibility on the same scale, and when the two don’t match, we have a problem”. Corporate control, he explained, means that AI policy stretches beyond existing national borders. He continued to praise this summit for bringing together leaders and representatives at the forefront of the discussion.

 Speakers called on the UK government to establish legislation, recalling Labour’s manifesto commitment to an AI Bill of Human Rights. The panellists broadly agreed on the importance of human rights, but appeared less aligned on what enforcement should look like in practice.

 In an exclusive interview, Michael O’Flaherty, the Council of Europe’s Commissioner for Human Rights, told Cherwell that political short-termism Continued on pg. 2 risks undermining efforts to address the long-term consequences of AI. “We are infected across our societies with shorttermist responses to the great issue of our societies and politics,” he said, adding that electoral cycles make it difficult for politicians “to think beyond the next election”. 

O’Flaherty also noted that younger generations no longer take the international human rights system for granted. “There is no longer an assumption that the human rights system is safe, that it will survive,” he told Cherwell. He argued that political leaders increasingly treat human rights selectively, saying: “Increasingly, leaders are treating human rights, if they are respected at all, as an a la carte buffet. We take those rights because they serve our goal. We’ll ignore others because they get in the way.

The summit also heard concerns about the effects of AI-generated misinformation on communities that may have limited access to, or awareness of, the technology itself. An audience question asked human rights experts to consider the impacts of AI on populations, such as those in India, who might be unaware of its existence. The focus on the impact of AI on populations in developing countries more vulnerable to the power of big tech was a key theme of the summit. 

Eileen Carter from the South African Human Rights Commission warned that underrepresentation of African languages in large language models could exacerbate existing inequalities. She also raised concerns about the effects of misinformation and deepfakes on elections, referring to more than 360 deaths associated with unrest and the amplification of narratives. 

The third panel, which offered a civil society perspective on promoting human rights in AI systems, offered some hope. Across all speakers, the potential for AI to transform the beliefs that underpin human rights, including the rights to water, education, and an interpreter, was debated. When asked about AI’s potential to unravel and alter the definition of human rights, such as what is included in the ‘right to education’ O’Flaherty told Cherwell: “The damage that can be done by AI has hit more dramatically and more visibly than perhaps any other sector within education through the phenomenon of cognitive loss, of cognitive transfer. We’re only beginning to realise the extent to which we have allowed education to be fundamentally undermined for an entire generation. 

“It’s time to seize that now and engage it. I’m not sure that the classic understanding of the human right to education is particularly well-framed for this extraordinary development.”

 In a discussion on different countries’ regulatory approach to AI, Suresh Venkatasubramanian, professor at Brown University and former AI advisor to US President Joe Biden, argued that AI systems should not be treated as inevitable developments beyond human control. “We continue, I think, to talk about AI as a thing that emerged out of nowhere that we just have to deal with,” he said. “We forget that these aren’t coming from an alien planet and they were built by entities that choose to build them in a certain way.” He called for pressure on technology companies to make different choices about how AI systems are developed. 

Questions of accountability also emerged during discussions about AI’s potential to improve access to justice. Research presented by Philippa Webb said that only 10% of people who have access to a lawyer can get one while there are thousands of bookings for an interpreter in a court covering 150 languages. However, this progress is not cemented as “general-purpose chatbots hallucinate 58% of the time on a legal task”. The use of AI in moderating legal proceedings raised fundamental questions about humanity and authority. Thompson Chengeta from Liverpool John Moores University said that “human rights is the last line of defence so when we say we want a human rights based governance… human rights is not an idea, it is an obligation. There is a gap that can only be recovered by a legally binding instrument”.

The event culminated in a panel offering an industry perspective. Co-host Dr Caroline Emmer De Albuquerque Green questioned representatives from Meta, Google, Microsoft, and OpenAI. Michael Karimian, who works in digital diplomacy at Microsoft, discussed the role of private sectors in engaging with AI, defining corporate responsibility as something which “provides transparency and identified risks”. 

All companies supported having formal programmes to ensure that they have an AI policy in place. Within their own companies, they suggested protective measures such as specialised training to ensure that policies, such as whistleblowing, are understood. Alexandria Walden from Google described how consultations with stakeholders could be used to identify when the product breached users’ boundaries. These assessments, according to OpenAI’s spokesperson, “feedback into tangible technical input” and “benchmark it”.

Fundamentally, understanding the rationale behind human rights is imperative, though lacking. Alexandria Walden from Google said: “Whether or not they understand human rights, they care enough to infuse it in their company… we might not always get it right, but it is something we care about.” 

Where former panels discussed controls to ensure human rights are considered in the development and implementation of AI products, this panel focused on mitigation. A spokesperson from Google pointed out that technology is inevitably going to move faster, so multiple tools are needed. Companies, OpenAI’s representative noted, still encounter issues such as discrimination and free speech. Consequently, they regard this as an additional policy which companies have a structured model to deal with, but newer companies, and newer industries, haven’t learnt it yet. 

In her concluding remarks, Green called for policy makers to “make the abstract actionable” at both governmental and corporate levels. She argued that, by framing the issues posed by AI as human rights issues, we gain a “foundation” for recognising AI’s many societal impacts.