UN's first global AI dialogue in Geneva faces three traps

A veteran internet policy hand has laid out a plain test for the United Nations' first Global Dialogue on Artificial Intelligence, which opens at the Palais des Nations on July 6-7. Writing in Tech Policy Press, Konstantinos Komaitis, a former Senior Director at the Internet Society, argues the meeting's success will be judged less by any formal output than by whether governments are still willing to talk to each other about a technology now viewed through the lens of strategic competition.

His starting point is bleak: "Never has international cooperation on AI been more urgent; yet rarely have the geopolitical conditions for cooperation been less favorable." He reaches for a Cold War analogy. Geneva, he notes, is where Reagan and Gorbachev first met in 1985, when the Cold War was, in his phrasing, "entering one of its most dangerous phases," and where dialogue began before trust existed. The programme this week is being pitched in a similar register.

Against that backdrop he flags three traps organizers should resist. The first is to avoid the security conversation entirely, which would leave the UN track sidelined while the real bargaining happens between capitals treating "frontier models, advanced semiconductors, compute capacity, cloud infrastructure and data ecosystems as instruments of national power." The second is the mirror failure: letting security swallow the whole discussion, which he argues would fragment global governance into competing blocs and crowd out the more ordinary cooperation the technology still needs. The third, and the one he pushes hardest, is treating the Global South as terrain to be contested rather than as an equal participant whose priorities shape the agenda.

The honest caveat is that this is an argument, not reporting. Komaitis does not walk through the delegations, does not name specific ministers, and does not predict what concrete follow-up mechanism, if any, will come out of the two days. He is telling readers how to read the meeting, not what it will decide.

What is worth watching is whether Geneva can hold as a venue where rival governments keep talking about AI even as export controls, compute policy and model access harden along national lines. That is a lower bar than a treaty, and for smaller economies and civil society groups it may be the more useful one.