In November, UK prime minister Rishi Sunak will host a summit to try to reach a global consensus on . While some people, such as tech entrepreneur , seem focused on the existential risk that AI might present, research indicates that some more prosaic and pressing aspects of regulating AI are being overlooked. Will global leaders be focusing on the right issues?
To assess the current state of AI regulation, at the Pontifical Catholic University of Rio Grande do Sul, Brazil, and his colleagues have analysed 200 public documents from 37 countries relating to AI ethics and governance, published between 2014 and 2022.
They looked at how frequently key concepts were mentioned and found that such guidelines focus mostly on transparency, security, justice, privacy and accountability.聽Transparency was most commonly mentioned, with the word appearing in 82.5 per cent of the documents. Security and justice appeared in 78 per cent and 75.5 per cent of the documents respectively.
Advertisement
But the work also shows that these AI guidelines tend to overlook labour and child rights, truthfulness and intellectual property protection. Only 1 in 5 documents mentioned labour rights, which are at the centre of the ongoing Hollywood writer鈥檚 strike in the US, just one of several hot-button issues that have emerged since OpenAI released ChatGPT in October 2022, leading to a rapid rise in the use of AI tools to generate text and pictures.
About 1 in 12 documents highlighted truthfulness as a concern, while intellectual property was mentioned in just 7 per cent of documents. Child and adolescent rights were present in just 6 per cent of documents.
, chief ethics scientist at AI company Hugging Face, and a former head of AI ethics at Google, says the research also shows that there is a lack of representation from a wide proportion of society: two-thirds of named authors of policy documents or suggestions were male, while a similar proportion of documents came from western Europe or North America.
The study highlights the challenges of regulating such a widespread technology like AI at a country-by-country level.
When it comes to the UK conference on AI, which will be held at Bletchley Park, where the enigma code was cracked in the second world war, 鈥渟uccess would be shared agreement on an accurate description of risk, as well as international consensus on shared understanding of risk, much of which was set out by the White House鈥�, says at the Tony Blair Institute for Global Change, a UK think tank.
Finding agreement and solutions in this way is important, if the analysis of the existing literature is anything to go by. The documents analysed by Santos and his colleagues offered plenty of suggestions, but just 2 per cent of guidelines provided practical methods to implement their recommendations. 鈥淚t鈥檚 mostly voluntary commitments that say, 鈥榯hese are some principles that we hold important鈥�, but they lack practical implementations and legal requirements鈥�, says Santos.
One document that is moving from simply suggesting towards actually making a difference is the European Union鈥檚 AI Act. Speaking at an event on responsible AI at the University of Amsterdam in the Netherlands on 10 October, , a member of the European Parliament and the co-rapporteur of the act, said that the legislation was at a 鈥渃rucial phase鈥�.
Benifei said 鈥渢here is a lot to do and the issues around fundamental rights will be crucial鈥�. His approach was supported by others at the Amsterdam conference. 鈥淲e鈥檝e come a long way, but we鈥檙e not there yet when we talk about responsible AI,鈥� says , president of ALLAI, a Dutch organisation dedicated to the development of responsible AI, which organised the event.
Given the disparate goals that various policy documents put forward, those who are looking at how to regulate the emerging technology are making the case for shared targets more strongly. 鈥淲e need to work together,鈥� says Benifei.
Patterns
