Recent AI capability jumps and an ungoverned race make one scenario increasingly likely: events force the US President — his Super Intelligence Force and a dozen US AI policy powerbrokers — to agree and enact, at wartime pace, national and international actions against extreme, global and imminent risks from AI: safety, national security, concentration of power, proliferation, including loss of control.

The first response would take weeks or months: sweeping presidential executive actions, in strict coordination with emergency, temporary executive agreements among a few big powers, to mitigate the gravest risks.

In parallel, extremely high-bandwidth US-China negotiations would be jump-started to design an extraordinarily bold, multi-phased, US-China-led treaty-making process. Its aims: (1) confidence measures and commitments within weeks or a few months, so the emergency agreements can hold; and (2) in stages, over months, the bold international institutions needed to prevent the worst safety, national security and concentration-of-power risks of AI — and to ground the astounding potential of advanced, human-controlled AI.

These closed-door consultations prepare the Administration and those powerbrokers for that process: to maximize the quality of the US position and the consensus behind it among key US AI policy powerbrokers. With the midterms behind us, Congress will ultimately have to approve any deal — and a deal China will trust must outlast two years.

What. One draft, many hands. In small closed-door sessions, the offices that shape US AI policy stress-test and co-edit a single plan for a US-China-led treaty process under emergency scenarios — the core do’s and don’ts that all or most agree on: phases and sequence, participants roles and vote weighting, trustworthy enforcement, and which clauses need supermajorities. Participants contribute evidence, objections and red lines; the draft records both agreement and unresolved differences. We convene and facilitate; participants shape the options; we remain responsible for the accuracy and quality of our analysis.

Who. Senior staff, advisors and envoys of executive-branch officials, frontier labs, Congress, and the political, media and faith leaders who shape the President’s AI policy and relevant multi-disciplinary independent experts — plus highly-informed and resilient citizen juries (as budget allows) since citizens trust neither government nor the labs. Mission-aligned organizations may join as partners, sponsors or co-hosts.

When. Provisional, to be confirmed: one or more of December 9 (Washington), December 13 (Miami, on the eve of the G20), January 13 (Washington) and February 10 (Washington, after the Force reports). Some sessions are open to remote participants.

What's discussed. Seven themes, each an open question — the spine of the conversation, not its answers:

●      The call — what would force the President's hand, and what the first decisions should look like.

●      The sequence — which first steps, what triggers the next, and how to scale fast without losing control.

●      Enforcement — what makes verification "trustworthy enough" for most stakeholders.

●      Bandwidth — the channels, tools and people a treaty needs from day one to scale.

●      The treaty's own risks — net benefit, concentration of power, sovereignty, and their guards.

●      Markets and citizens — what the economy can live with, and what share citizens get.

●      AI consciousness — foster a head-on debate, understanding, convergence, compromise and win-wins on key aspects that are increasingly crucial to policy

Aims, context and the provisional program: see the summary of The Deal of the Century on AI.

Take part. One email is enough: rufo@cbpai.org

●      I'd like to take part, as a delegate or an expert.

●      Someone should be in the room — here is a name, and who could make the introduction.

●      We'd like to partner or co-host — think tanks and organizations: rufo@cbpai.org, with times for a 20-minute call.

Further details

The missing lane. Washington has plenty of convenings on guardrails, standards and China. Few ask: if Trump and Xi were to move — or be moved — beyond protocols toward a bold treaty-making process, what process should the US aim for, and can a critical mass of US AI powerbrokers converge on a plan to steer it to success? That is the lane — Amodei's step two, "democratic coordination," widened from the labs to everyone who shapes the President's AI policy. The September 29 Accord began the labs' coordination; without China, and without the officials and voices who will decide what a treaty says.

Shared goals, different predictions. As in the 1946 Acheson–Lilienthal consultations, we assume contributors largely share humanist goals and diverge on key AI predictions: whether China wins the race, whether superintelligence stays under human control, whether verification can work, whether a treaty can prevent both runaway AI and AI-enabled authoritarianism. The evidence on each is moving fast, and one key figure revising one prediction can reshape a consensus.

How. Each session works on the latest draft: what's wrong, what's missing, what their principal could live with. Structured methods — Delphi pre-surveys, structured disagreement mapping, adversarial collaboration — to record agreement, persistent disagreement and what evidence could change a position. Each contribution lowers the bar for the next: once two or three offices have shaped the text, the fourth joins something already respectable.

What it produces. No public statement, declaration or signatory list. Three private outputs: a better memo — evidence and objections considered in Strategic Memo v3.0, attributed by name only with consent; a coalition map — an anonymized account of convergence and disagreement, with named positions shared only by explicit permission; and, if enough converge, a coalition ready to ask the President for a meeting together. If not, the memo stays on file with the Administration, ready when he reaches for it.

Preparing the US side for dialogue. In early 1946, the Acheson–Lilienthal consultations let Washington align internally, months before it took a proposal for the international control of nuclear weapons to the UN. Same aim here — driven by NGOs, with key national stakeholders, toward a proposal the other side can accept.

Where to go next.

• For more information on the program, contact us at rufo@cbpai.org

• The Deal of the Century on AI — the aims and the provisional program. Read before you come.

• Strategic Memo — v2.6, 356 pages; the v2.9 update; v3.0, about 300 pages, due November 20.

Last updated November 6, 2026