Eli Dourado, who has served as a State Department advisor and participated in multiple treaty negotiations, as well as served on technical committees for UN technical agencies, offers a critical assessment of how international governance efforts actually function in practice. His critique comes at a moment when multiple international AI governance initiatives are actively taking shape—the UN General Assembly adopted its first resolution on AI in March 2024, the UK hosted the AI Safety Summit at Bletchley Park in late 2023, the G7 launched the Hiroshima AI Process, and the EU finalized its AI Act—making the question of whether such frameworks can deliver meaningful oversight increasingly urgent.

In a post on X, Dourado questions the reassurance that many find in the phrase "international efforts" when it comes to addressing governance challenges, including those surrounding AI. Drawing on his firsthand experience, he describes the international sector as "unbelievably dysfunctional."

According to Dourado, every treaty or international agreement becomes an opportunity for each participant to manipulate the broader policy environment. He notes that participants frequently have little interest in the actual subject of the agreement and instead seek to use it as leverage for unrelated objectives. He has observed countries use their limited leverage in multilateral agreements to undermine American industry, circumvent sanctions, manufacture a pretense of international legitimacy for domestic illiberal policies, or push technology development in an authoritarian direction.

The damage from this dysfunction, Dourado argues, is contained by one key factor: there are essentially zero major countries—and not many smaller ones—that will genuinely bind themselves to obligations they do not narrowly want. When a previously signed treaty becomes inconvenient, it is often quietly ignored or reinterpreted. This dynamic has precedent in other domains—the Intermediate-Range Nuclear Forces Treaty collapsed in 2019 after years of alleged violations, and compliance disputes have plagued agreements from the WTO to the Chemical Weapons Convention.

Dourado also highlights the composition of treaty delegations handling industrial issues, noting that they tend to reflect the full spectrum of special interests at play. A US AI treaty delegation, for example, might be led by a State Department ambassador but would include advisors from across the interagency, major AI labs, other large technology companies, tech investors, and civil society representatives. The parallel to other regulatory arenas is notable: in telecommunications, the ITU's standard-setting processes have long been contested ground where authoritarian governments push for greater control over internet infrastructure, while industry representatives advocate for positions aligned with their commercial interests.

"International efforts," Dourado concludes, is therefore not a reassuring answer to the governance problem. It is, rather, the name of another enormous and unresolved governance problem in itself.

Source: Marginal Revolution