Standing Global Citizens’ Assembly on AI

Source

• Article: Let The People Set The Pace Of Frontier AI. By Hélène Landemore and Audrey Tang. Noema magazine, 23 September 2026

URL = https://www.noemamag.com/let-the-people-decide-the-pace-of-frontier-ai/

Discussion

DeepSeek Synthesis based on the Noema article by Hélène Landemore and Audrey Tang, prompted by Michel Bauwens.

The authors propose creating a standing global citizens’ assembly of 1,000 people chosen by lot from around the world, which they call the Global Citizens’ Assembly on AI. This body is designed to act as a credible, independent witness and assurance mechanism to help humanity set the pace and limits of frontier AI, breaking the current collective-action deadlock.

Purpose and Mandate

The assembly would be charged with helping to set the purposes and limits of powerful AI and with receiving the evaluators’ findings on whether those limits are being kept. It would address fundamental questions on behalf of humanity, such as:

What must AI never be allowed to do?

What do we want it to do for us and for future generations?

Who should govern it, and by which principles?

Composition and Operation

Selection by Lot: Members are chosen by lottery from every continent, not through national elections. The authors argue that a person’s stake in the AI race is not conditional on knowledge, credentials, or nationality.

Size (1,000 Members): The number is chosen for symbolism and salience. A body the size of a large parliament cannot be easily ignored; a lab leader who refuses to answer a thousand people drawn from every continent must do so publicly on the planetary stage.

Continuous Sitting: The assembly would sit continuously rather than produce a single report. Members would serve fixed, part-time terms, with a third of the seats regularly redrawn by lot so the body keeps its memory without hardening into a class of its own.

Deliberative Process: Members would deliberate over months, both online and with local support. Their work would be facilitated by AI tools for translation, sorting contributions, and identifying points of agreement and disagreement, while members retain the authority to judge summaries and set the agenda.

Support and Protection: Members would receive financial compensation, translation and interpretation, secure communication, and protection for those whose governments may dislike what they say.

Independence and Accountability

Independent Civic Secretariat: An independent secretariat would pool disclosed funding from participating governments and labs without letting funders influence recruitment or tamper with evidence.

Duty to Answer: Each recommendation from the assembly would go to a named institution (a lab or government) with a deadline to respond in public: stating what it accepts, what it rejects, and why. The authors stress that this “duty to answer” is the key to turning moral authority into political legitimacy.

Rationale: Why This Body Is Necessary

The authors argue that the current AI race is a “stag hunt”—a collective-action problem where all parties would benefit from mutual restraint, but each fears being the sole defector. The way out is not coercion but an assurance mechanism requiring verification and a credible witness that no lab or state controls. The UN is unsuitable because it is structurally entangled with the rival powers and speaks only for nation-states, not humanity. Humanity itself has the strongest claim to credibility on this issue, as its stake is universal and undiluted by any single national or corporate interest.

Implementation Strategy: “Start with the Willing”

The authors outline a pragmatic, phased approach:

Coalition Funding: A small coalition of governments and foundations can fund the assembly through the independent civic secretariat, creating a focal point no lab or state controls.

Lab Participation: Labs could open their systems to evaluators vetted under rules the assembly sets, route findings to the assembly, and accept a public duty to answer its recommendations within a fixed deadline.

Legal Anchor (EU): A clause in a revised EU AI code of practice could oblige signatories to answer the assembly’s recommendations in public, giving the duty to answer its first legal anchor in a market no frontier lab can afford to leave.

Measurable Pacing: Participating labs could pilot a transparent safety-research budget of 5% to 20% of their compute under published accounting rules, giving evaluators concrete metrics to verify.

The authors conclude that “the racers cannot decide that pace alone; they need a witness they all trust. Let the people be that witness: Give them a seat, a say in the terms and a right to an answer.”.