UN Begins Full-Scale Discussions on Regulating Military AI with New International Forum

The United Nations is launching full-scale discussions to regulate the military application of artificial intelligence (AI).

According to ABC News on the 20th local time, the United Nations Institute for Disarmament Research (UNIDIR) is set to launch its first international forum to discuss the risks and control measures of military AI.

UNIDIR, located in Geneva, Switzerland, established the "Center of Excellence for AI, Peace, and Security" this past summer, which is dedicated to issues surrounding military AI and international security.

Centered around this hub, the international forum launching on the 22nd will serve to bring together government officials, defense contractors, software developers, civil society, and academic experts to discuss the issue of controlling military AI.

Additionally, the UN is pushing to establish voluntary codes of conduct for companies participating in the development of military AI.

The goal is to set behavioral standards for the industry to ensure compliance with international law and human rights even during the development process of military AI, thereby providing autonomous guidelines.

While discussions to establish international norms for military AI have previously stalled, the UN believes that it can no longer delay its response.

"We can no longer leave the urgent issue of AI in the military domain in the shadows," said Robin Geiss, Director of UNIDIR. "It is now time to pool our capabilities and directly address the issue of military AI."

Geiss pointed out that as AI is rapidly adopted across various military sectors such as early warning systems, intelligence collection, and reconnaissance, it is becoming increasingly difficult to even determine when humans might lose practical control.

Meanwhile, the UN also plans to initiate separate discussions to counter the risk of armed groups misusing AI.

(Photo: Yonhap News)
※ Please note: This article was translated by AI and may contain errors.