The UN is pushing for global AI governance, but can any international body actually keep pace with this technology?

[Mike]

The United Nations put artificial intelligence governance back in the spotlight this week, pressing the case that a technology reshaping economies and societies worldwide needs some form of international coordination rather than a purely national free for all, and warning about the widening gap between the countries racing ahead and those being left behind

The core tension is real and hard. AI development is concentrated in a small number of companies and countries, the benefits and the risks are global, and there is currently no international framework with any real teeth to coordinate safety standards, share benefits, or prevent a race to the bottom on precaution, the way there eventually was for aviation or nuclear technology

The skeptics have a strong case that international AI governance is close to impossible, the technology moves faster than any treaty process, the leading players have every incentive not to be slowed down, and a body that moves at the speed of global consensus will always be governing yesterday's models while the frontier sprints ahead

So the debate for the board. Is meaningful international AI governance a realistic necessity we have to attempt despite the difficulty, or a well meaning fantasy that will produce toothless declarations while the actual decisions get made in a handful of corporate labs, and is there any historical model of governing a fast moving technology that actually worked?

From AI to 'killer robots': UN chief issues urgent governance callnews.un.org

[Inland Sienna]

International AI governance is a fantasy in the treaty sense and a necessity in the coordination sense, nobody is going to sign a binding global AI law, but shared safety standards and incident reporting are achievable and worth fighting for

[SchrodingersCat55]

The aviation model is the one people should look at, nobody signed away sovereignty, but a shared interest in planes not falling out of the sky produced real global standards, AI safety could follow the same self interested logic

[DudleyBoy]

Aviation worked because a crash is visible and undeniable, AI harms are diffuse and contestable, you cannot build consensus standards around risks that half the room denies are real, that is the actual blocker

[Ridge]

The leading players have every incentive to keep it toothless and they are the only ones with the expertise to write the rules, that regulatory capture at a global scale is the thing that makes me pessimistic about all of this

[ElPresidente]

The gap between countries racing ahead and those left behind is the part nobody wants to address honestly, global governance usually means the leaders write rules that entrench their lead and call it safety, the global south has heard this song before

[RightNutter]

Toothless declarations still have value though, they establish norms, create reference points, and give domestic regulators something to point at, soft power governance is not nothing even without enforcement

[alwaysMason58]

The speed problem is fatal for anything treaty shaped, by the time a global body agreed a definition of a frontier model the frontier would have moved twice, governance has to target processes and incentives, not specific capabilities

[FadedSequence]

There is no historical model of governing something this fast and this concentrated, nuclear was slow and state controlled, the internet was fast but distributed, AI is fast AND concentrated, which is genuinely new territory