China opts out as 60 countries agree on the responsible military use of AI
Around 60 countries have agreed on a blueprint for the responsible military use of AI – but China was not one of them.
The Responsible AI in the Military Domain (REAIM) summit took place in South Korea, with the UK, US, France and Germany among those nations who backed the legally non-binding document.
However, China was one of the 30 countries that sent a representative to Seoul but did not support the document.
Co-hosted by the Netherlands, Singapore, Kenya and the United Kingdom, the summit is the second of its kind – with the first held in The Hague last year.
Professor Stuart Russell, one of the world's leading experts on AI, explained the type of situations the blueprint looks to avoid.
"If you had autonomous ships patrolling international waters and one of the ships perceives that a ship from another country is threatening, then it might respond, it might retaliate and then you have a real war," he said.
"That can escalate very quickly and then it may be impossible to reverse that."
At last year's event, 60 nations, including China, endorsed a modest call to action without legal commitment.
Ruben Bekelmans, the Netherlands defence minister, told Reuters this year was about "getting more action".
This, he said, included what risk assessments should be made, what conditions are important, such as human control, and how to manage risks.
Other representatives also said the blueprint was aligned with discussions and developments of AI use in the military – such as AI-enabled drones being used by Ukraine, which also supported the document.
The document also outlined the need to stop AI from being used to create weapons of mass destruction for terrorist groups.
Additionally, the importance of human control remains and ruling out AI's involvement in nuclear weapons employment.
The summit aims to ensure the discussions are not led or dominated by a single nation or entity.
However with 30 nations, including China, not supporting the document, it is clear there are stark differences in opinion on AI.
Mr Brekelmans said the representatives "need to be realistic that we will never have the whole world on board".
"How do we deal with the fact that not everyone is complying? That is a complicated dilemma that we should also put on the table," he added.
Architects of the summit will now take the action plan to the United Nations in a bid to have it more universally adopted.
Giacomo Persi Paoli, head of programme security and technology at the United Nations Institute for Disarmament Research (UNIDIR), said any risks should be mitigated by countries speaking to each other between summits.
"The blueprint is an incremental step forward," he said. "By going too fast, too soon, there is a very high risk that many countries do not want to engage."
Related topics
Join Our Newsletter
