The agreement was introduced at the Responsible AI in the Military Domain (REAIM) summit in Seoul on Tuesday (Sept. 10), attended by representatives from over 100 countries, including the United States.

Though not legally binding, the agreement seeks to ensure that "human control and involvement are maintained in all decisions related to nuclear weapons deployment."

It also states that "AI applications should be ethical and human-centric."

South Korean Defence Minister Kim Yong-Hyun described AI as a "double-edged sword," noting, "while AI enhances military operational capabilities significantly, it also poses risks if misused."

The summit's declaration did not outline any sanctions or penalties for breaches.

It acknowledged the need for substantial progress to keep pace with advancements in military AI and called for ongoing discussions to develop clear policies and procedures.

The Seoul summit, co-hosted by Britain, the Netherlands, Singapore, and Kenya, builds on the inaugural event held in The Hague last February.

It aims to be the "most comprehensive and inclusive platform for addressing AI in the military domain."