The United Nations (UN) High-level Advisory Body on Artificial Intelligence published, on Friday 20 September, its final report on the global governance of artificial intelligence, soberly entitled ‘Governing AI for Humanity’.
Presented just before the opening of the ‘Summit of the Future’ in New York on Sunday 22 September, the document advocates greater UN involvement in the regulation of AI.
It highlights the contradictory challenges of establishing governance for a technology that is developing so rapidly and is the subject of massive investment.
The report points out the paradox of the AI framework: despite the “hundreds of guides, frameworks and principles (that) have been adopted by governments, companies and consortiums, and regional and international organisations (...), there is a global governance deficit in AI”.
While offering major advantages for the development of countries and populations, AI remains a tool with a fairly high potential for danger: amplification of discrimination, massive disinformation, capacity of massification of problems or cause controversy, uncertainty about its military use, difficulties in legislating and organising international coherence, with a disparate set of overlapping or telescoping initiatives.
The EU is not to be outdone, having itself produced numerous frameworks on AI, as well as the high-profile AI Act, the first direct regulation on artificial intelligence.
According to the UN report, there is “a mosaic of norms and institutions (that) is still in its infancy and full of gaps. Accountability is often conspicuous by its absence, including for the deployment of non-explainable AI systems that have an impact on others. Compliance is often voluntary”.
The document therefore proposes setting up a scientific body for international governance, to ensure a global and coherent approach to AI.
This body could be responsible for receiving and investigating reports of incidents or abuses, reporting on serious violations, verifying compliance with international obligations and developing global standards for AI safety.
To see the report: https://aeur.eu/f/djf (Original version in French by Isalia Stieffatre)