Article 8

Gates’s final proposal is his most ambitious: a new framework for global AI governance, comparable to the reorganization of the U.S. government after 9/11, but much broader. Yet Gates acknowledges that politics moves slowly while AI moves fast. The result is a tragic mismatch. Can we imagine governance that does not depend on states, but instead emerges from below—from civil society and technical communities?

Gates concludes his manifesto with a call for global action: “This unprecedented technology demands an unprecedented global response.” He proposes creating national institutions to coordinate AI policy across different ministries (labor, security, education, energy, and so on) and, in parallel, an international organization that would serve as an equivalent of the IAEA for AI.

The limits of the IAEA analogy

The analogy is apt, but it also reveals its limits. The IAEA was created during the Cold War, when two superpowers shared a minimal interest in avoiding mutual annihilation. That shared interest made some agreements possible. Today, rivalry between the United States and China is systemic, not merely ideological, and spans every domain. There is no clear shared interest in slowing AI; in fact, each sees the AI race as a zero-sum game in which whoever arrives first controls the future.

The geopolitics of AI.

Gates mentions the need for cooperation between the United States and China, but does not explain how to overcome the obstacles. The former sees AI as its technological and military advantage; the latter sees it as an opportunity to surpass the West. Both are investing hundreds of billions in AI development, and both have autonomous weapons programs. In this context, a non-proliferation agreement seems unlikely.

Moreover, European Union countries, which could act as a regulatory “third way,” have less technological capacity and are more focused on regulation (the much-discussed risk-based approach). But without technological power, regulation can be ignored or circumvented. The EU can fine big tech companies, but it cannot stop them from developing more advanced models elsewhere.

The innovation race as an obstacle.

Gates acknowledges that “geopolitical and economic incentives are pushing everyone at full speed.” This is the reality that undermines any proposal to slow down. The logic of platform capitalism, combined with the logic of national security, creates a vicious circle: each country fears the other will pull ahead, so it accelerates. And when everyone accelerates, the result is a coordination disaster.

Possible ways out of the acceleration trap

The only way out of that trap is a technological “peace agreement,” similar in some respects to the one that ended the nuclear arms race. But that agreement took decades of crises, and the risk posed by AI is more diffuse and less visible (a nuclear war is immediate; AI’s harm is incremental and systemic).

Political imagination: governance from below.

If state and global governance are unlikely, perhaps the solution will come not from above but from below. The history of the internet shows that technical standards and ethical protocols can emerge from communities of practice before states act. TCP/IP, the protocol that makes the internet work, was created by researchers and prevailed because it was useful, not because a government decreed it.

Similarly, “ethical AI” standards could emerge from universities, NGOs, and responsible companies. A “trusted AI” seal could have the same effect as environmental certifications. If consumers and investors begin demanding transparency and fairness, companies will face pressure to comply, even without a law.

But this approach has a limit: malicious actors and authoritarian states will not submit to voluntary standards. AI is a dual-use weapon, and anyone who wants to use it for harm will not ask permission.

The role of civil society.

At the end of his manifesto, Gates calls for a wider circle of participants: workers, students, community leaders, religious leaders, parents, and educators. This acknowledges that AI governance cannot be the sole concern of technocrats and CEOs. But it sounds like a plea, a wish more than a program.

Civil society needs to organize, apply pressure, and mobilize. In recent history, the climate movement has managed to put the issue on the agenda, even when governments were reluctant. A movement for “AI justice” could have a similar effect, demanding moratoria, transparency, and democratic participation in the design of these systems.

Time as the enemy.

Gates repeats it several times: we do not have the luxury of moving slowly. AI is advancing exponentially, while institutions move linearly. The gap between the pace of technology and the pace of politics is the real threat. By the time we have created a regulatory framework, technology will already have outgrown it.

Regulation that keeps pace with AI

This temporal asymmetry calls for “adaptive regulation”: frameworks that do not try to regulate the technology itself, but its impacts, and that can be updated quickly through agile procedures. Something like a regulatory “fire department” that responds to fires in real time.

Final conclusion: the clarity of powerlessness.

Gates’s manifesto is a document of clarity and powerlessness. It sees the dangers, contradictions, and urgency clearly, but lacks a real political lever to change them. His proposal for new global governance collides with geopolitical reality; his fiscal proposal with entrenched interests; his proposal for “human reserves” with the market.

Still, clarity is no small thing. In a world where most leaders deny or trivialize the risks, there is incalculable value in Bill Gates standing up and saying, “This is serious.” His manifesto is a mirror reflecting our collective inability to act. The question it leaves us with is not “What will Gates do?” but “What will we do?”

Because AI does not wait, and history does not write itself.

Bibliography for article 8

  • Gates, B. (2026). The turbulent AI era is here… Gates Notes.
  • Kissinger, H., Schmidt, E. & Huttenlocher, D. (2021). The Age of AI: And Our Human Future. Little, Brown.
  • Sanger, D. (2023). New Cold Wars. Crown.
  • Nye, J. (2024). “AI and American Power.” Foreign Affairs.
  • Lessig, L. (2006). Code: Version 2.0. Basic Books.
  • Ostrom, E. (1990). Governing the Commons. Cambridge University Press.
  • Floridi, L. (2023). The Ethics of Artificial Intelligence. Oxford University Press.