Great-power dialogues on Artificial Intelligence are necessary but insufficient for global AI governance. Discuss.

The recent U.S.–China dialogue on AI, which comes with an incidents channel, shows that AI has become part of how great powers manage crises. AI-enabling goods made up only one-sixth of world trade in 2025 but accounted for 42% of its growth [1]. Talks between the two leaders are therefore necessary. However, a G-2 cannot govern a technology that the whole world uses.

Why great-power dialogues are necessary

• Crisis prevention: the two leading AI powers are also nuclear rivals. An incidents channel reduces the chance that an AI-enabled military or cyber mishap escalates, much like the Cold War hotlines did.
• Concentration of capability: most advanced chips, compute and frontier models are in these two countries. Rules made without them cannot be enforced.
• Slowing an unregulated race: regular dialogue builds trust and makes competitive corner-cutting on safety less likely.

Why they are insufficient

• Exclusion: a two-member forum risks turning the Global South into rule-takers of standards it never helped shape.
• Narrow agenda: security-focused talks leave out bias, jobs, data rights, deepfakes and energy use. These are the issues covered by the UNESCO Recommendation on the Ethics of AI, adopted by 193 states [2].
• Fragility: the talks are non-binding and depend on personal rapport between leaders. Rivalry over chip export controls continues, so progress could stall in the next trade dispute.
• Bypassing multilateralism: UNGA Resolution 79/325 (2025) created an Independent International Scientific Panel on AI and a Global Dialogue on AI Governance, which first met in Geneva in July 2026 [3]. G-2 deals can weaken this inclusive process.

Way forward

• Link bilateral and multilateral tracks: U.S.–China safety commitments should feed into the UN Global Dialogue.
• Speak for the Global South: India's New Delhi Declaration on AI Impact, endorsed by 92 countries and international organisations, shows that inclusive consensus is possible [4].
• Build shared capacity: open compute, datasets and safety-testing tools can spread AI's benefits beyond a few countries.

Great-power dialogue can lower the risk of catastrophe, but it cannot make AI governance legitimate for everyone. A layered system is needed: a U.S.–China channel for managing crises, placed within UN-led and inclusive forums. Such a system would reflect India's principle of "Sarvajan Hitaya, Sarvajan Sukhaya" (welfare for all, happiness for all) [4] and help make AI a tool for the SDGs rather than a new source of global division.