Officials at the Responsible AI in the Military Domain (REAIM) summit in Seoul, which involved nearly 100 countries including the United States, China and Ukraine, adopted the “Blueprint for Action” after two days of talks.
The agreement — which is not legally binding, and was not signed by China — said it was essential to “maintain human control and involvement for all actions … concerning nuclear weapons employment”.
It added that AI capabilities in the military domain “must be applied in accordance with applicable national and international law”.
“AI applications should be ethical and human-centric.”
The Chinese embassy in Seoul did not immediately respond to a request for comment.
Militarily, AI is already used for reconnaissance, surveillance as well as analysis and in the future could be used to pick targets autonomously.
Russia was not invited to the summit due to its invasion of Ukraine.
The declaration did not outline what sanctions or other punishment would ensue in case of violations.
The declaration acknowledged there was a long way to go for states to keep pace with the development of AI in the military domain, noting they “need to engage in further discussions… for clear policies and procedures”.
The Seoul summit, co-hosted by Britain, the Netherlands, Singapore, and Kenya, follows the inaugural event held in The Hague in February last year.
It bills itself as the “most comprehensive and inclusive platform for AI in the military domain”.
