UN proposals for global AI fund do not go far enough, according to digital economy expert

This week, an artificial intelligence (AI) advisory board at the United Nations (UN) released a report, outlining seven recommendations for bridging the gaps in AI risk and governance. Among guidance on information sharing and policy dialogue, the report recommended the establishment of a new global AI fund designed to help less developed countries harness its powers and potential more readily.

Dr Itziar Castelló, Reader in Corporate Social Responsibility at Bayes Business School, said the report needed to focus more squarely on data protection and preventing AI from being misused across the world.

"AI models and platforms should be conceived as critical institutions of modern democracies,” she said.

“Serving as essential parts of the public sphere for deliberation and individual and collective decision making on matters of public concern.

“In such a conception, AI platforms should become tools for maintaining – rather than eroding – democratic culture and critical thinking. 

“The UN and western nations therefore need to go beyond the proposal of creating a global AI fund to prevent misuses related to the abuse of data.”

To remedy this, Dr Castelló believes there needs to be a rethink about what AI is and the purposes it serves.

“Firstly, authorities should raise awareness of the need for the re-conceptualisation of AI models and platforms as spaces that construct collective knowledge in society, through public debate. This means AI platforms should be perceived as institutions of a democratic society, rather than solely private or technology-driven platforms. 

“Secondly, I would advocate for the development of a global framework followed by regulation and codes of conduct aimed at safeguarding deliberative space at firm level.

"This means not only preserving privacy of data, transparency, safety and inclusion, as advocated by most technology perspectives, but also deeper intervention into how communication and knowledge is constructed and amplified within AI models and tools."