The first multilateral discussion on AI safety called for international cooperation while identifying concerns about AI's risks if used to gain hegemony.