CBPAI September Newsletter: In DC Sept 14-19 — the US-China AI Dialogue Is On, Its Scope Is Not

The first Trump-era US-China talks devoted solely to AI are being lined up for the coming weeks, ahead of President Xi's September 24 visit to Washington. Their agenda — cyberattacks, distillation, labs sharing threat data — is a tiny fraction of what Xi, the leading labs and most US voters have said the moment requires. Closing that gap is the niche we occupy. We'll be in Washington September 14–19 for the Pro-Human Assembly, meeting prospective co-hosts, co-steering partners and funders of our planned Q4  Roundtables for a US-China AI Treaty in DC. Come find us.

____________________________________________ 

The dialogue is on. Its scope is not.

Reuters reports that the first official US-China talks devoted exclusively to AI since President Trump returned to office — led by Treasury Secretary Bessent — are being prepared for mid-September, or October according to Treasury, days ahead of President Xi's September 24 summit in Washington. The proposed agenda: AI-directed cyberattacks, model distillation, and a US proposal that American and Chinese labs share threat information directly.

Serious analysts are setting the bar accordingly. Carnegie's Scott Singer writes in Foreign Policy that success would be "a serious second meeting and eventually a permanent high-level channel." Brookings argued in July that the two sides "should not aim at a binding treaty, or even shared norms."

We think that is the right first step and the wrong ceiling. In July, President Xi called to "constantly refine measures to forestall loss of control." Days later, a thousand AI scientists — with Anthropic's and OpenAI's support — requested that the US government back an "international effort" to pace the frontier. 63% of US voters expect humanity to lose control of AI.

Nobody in Washington is yet resourcing the question those statements pose: if the Administration decided tomorrow to scale the Bessent dialogue tenfold or more — in scope, bandwidth and speed — what machinery, safeguards and negotiating sequence would it need? In 1946, Stimson's advice to deal with Moscow directly was ignored, and the Baruch Plan died at the UN. This time the bilateral channel exists. What it lacks is a plan proportionate to the stakes — and a critical mass of people around the President asking for one. That gap is where we work.

The Deal of the Century Roundtables — Washington, Q4 2026

Two closed-door days under the Chatham House Rule for 20–30 staff, advisors and envoys of key "humans first" US AI policy powerbrokers — in the executive branch, the labs, and political, religious and media circles — plus the people best placed to brief them. No joint statement, no signatory list. The single product is a documented map of where they converge and diverge on the design of a proper treaty-making process — measured rather than asserted, through Delphi elicitation, opinion-cluster mapping and adversarial collaboration. Co-hosts and partners are covered by a Convening Protocol with unconditional withdrawal rights and no position attributed to anyone.

Five DC think tanks and advocacy organizations are in discussions with us as co-hosts or supporting partners; we expect two or three more. Dates and venue will be set with them, in Q4. The program and one-pager were updated on September 9.

Beyond co-hosting: a US-rooted consortium

Our coalition of ten small NGOs was born in Europe; the work is in Washington. So we are inviting two or three US advocacy organizations to co-steer the initiative — full decision rights on a new steering board, with the Roundtables as the consortium's first product rather than a one-off event.

If your organization wants a proper treaty on the President's agenda but lacks the bandwidth to build the executive-branch lane itself, this is the lane. It complements, and does not compete with, domestic regulation, technical standards or the Track 1.5 dialogues run by far larger institutions — whose leaders we intend to invite.

See us in Washington, September 14–19

We'll be in DC for the Pro-Human Assembly on September 15. CBPAI is a signatory of the Pro-Human AI Declaration and regards itself as the movement's diplomatic complement: the Declaration's first principle — humans stay in charge, no race to superintelligence — cannot be secured within one jurisdiction. Someone has to work the treaty lane toward the executive branch. That is us.

Between the 14th and the 19th we are holding slots for prospective co-hosts, co-steering partners, funders, and anyone who can introduce us to the staff or advisors of the powerbrokers named on the Roundtables page. Write to rufo@trustlesscomputing.org with two times that work for you.

Why fund this lane — a section to forward

AI-governance philanthropy funds three things well: domestic regulation, technical standards and verification, and Track 1.5/2 dialogues housed at well-resourced institutions. It barely funds the layer where scope is decided — the design of the treaty-making process itself, and the convergence of the dozen humans-first figures who can put a proper one on the President's desk. A plan for the plan.

Our record in that lane, on about $140,000 in lifetime donations: a 356-page Strategic Memo with 150 pages of powerbroker profiles; 90+ briefings with their staff and advisors; over 30 hours with senior officials or advisors reporting directly to four of them; a 43-page Executive Summary published in August, with Memo v3.0 due later this month. Our burn is about $7,500 a month; $500K–$1M would let us scale roughly tenfold within months, as laid out in the Executive Summary. Donations are US tax-deductible through our 501(c)(3) fiscal sponsor.

Calls to action

In order of priority:

• Meet us in DC, September 14–19 — send two proposed slots to rufo@trustlesscomputing.org.

• Apply as co-host, supporting partner or co-steering partner of the Q4 Roundtables: attach the Convening Protocol, as-is or amended, and propose a 20-minute call.

• Donate — or forward the section above to a fitting donor.

• Introduce us to staff, advisors or introducers of key US AI policy powerbrokers.

• Join us in other ways.