The Deal of
the Century
Roundtables
Washington D.C.
TBD Early Q4, 2026
Why This Matters
A declared race to Superintelligence is bringing humanity to a stark three-way fork: catastrophic loss of control, authoritarian capture, or humanity's triumph through a properly governed AI. Many AI scientists and leaders state that a point of no return is plausibly a few years or even months away.
A bold global treaty is the only way. A dozen individuals are in a unique position to support and guide Trump in co-leading with Xi a bold treaty for AI and in shaping history — as others did in 1946, leading Truman to formally propose history’s boldest treaty, the Baruch Plan, on the very day Trump was born.
They can succeed this time by leveraging modern state-of-the-art treaty enforcement and diplomatic communication socio-technical systems, and jointly creating more suitable, trustworthy, and mutually trusted ones at wartime speed. They have a much better chance, arguably, by avoiding Truman's crucial error of preferring an initial deal with allies and a UN process rather than with Stalin, and by incorporating resilient mechanisms for AI power and wealth sharing.
Why Attend
The roundtables will aim to foster greater convergence between President Trump and a critical mass of key US AI policy powerbrokers (i.e., influencers) on a radical and suitable shift and expansion of current US AI policy.
Amidst a strategic context that has changed more in the last two months than in the last three years, a historic, if daunting, blank slate has opened up for a bold, new US AI policy fitting to the new context.
The opening is unprecedented. Radical recent advances in AI offensive capabilities and shifts in public and elite sentiments have led the Administration to drop its long-standing handoff policy approach to AI safety and the concentration of power, and the informal post-humanist alliance that supported it, thereby swinging wide open the Overton Window of AI policy solutions.
While the hands-off approach proved suitable to date, without major safety accidents, and key to remaining ahead of China, many of the underlying risks and challenges that had built up suddenly gushed out.
The Administration found itself rushing to implement stop-gap solutions to critical safety risks and starting to explore widely-supported new proposals for the citizen sharing of AI power and wealth, including up to 50% citizen co-ownership of top AI labs. This is a huge opportunity and responsibility that befits a few key US AI policy powerbrokers: crafting with the Administration a durable new US AI policy, fitting for the new context, humanist, and cautiously techno-optimist.
In our view, such a new policy can and should achieve the following goals to be a political success:
Future-proof US economic leadership
Contain emerging military threats
Reduce AI power and wealth concentration
Prevent the enormous risks of loss of human control, misuse and superpowers' conflict
Realize AI's astounding potential for Americans
For more, read the Strategic Backgrounder chapter at the end of this page.
Towards a cohesive, widely shared new US AI policy
Amid the vast array of options and opinions, Trump needs a new AI policy that can reconcile the different needs and priorities of a critical mass of key powerbrokers — such as security agencies, AI lab leaders, ordinary citizens, and other AI policy powerbrokers — who can support and guide him.
Each brings a mix of different priorities such as dangerous proliferation, civil liberties, concentration of power or wealth, reliable banning of ASI, and avoidance of excessive regulation that could stifle the industry as it happened for nuclear energy.
These roundtables convene those powerbrokers — and the advisors and introducers who can reach them — to work through exactly these reconciliations, session by session.
Roundtable Participants
Invitations will start mid-July. Prospective participants include relevant advisors, staff, envoys and introducers to key formal and informal US AI policy powerbrokers.
Key among them, nowadays, are certain US officials (including Kratsios, Rubio, Vance, Bessent, NSA director Rudd, and WH Chief of Staff Wiles); political, religious and media figures (including Pope Leo XIV, Bannon, Carlson, Beck, DeSantis, and Rogan); and humanist AI lab leaders (including Amodei/Clark, Altman, Hassabis, Musk and Sutskever). Most of them were profiled in depth over 150 pages in our 356-page Strategic Memo of the Deal of the Century v2.6.
As of July 2026, we believe the best positioned to act as catalysts are Kratsios, Amodei/Clark, Altman, Rubio, Rudd, and Wiles.
Several DC-based NGOs and think tanks will be invited, as they can introduce, validate, and contribute, especially those respected by the IC and the Administration.
More Information
The Program chapter below for a detailed description of the joint answers we'll be pursuing.
The Backgrounder chapter below for a detailed rationale, basic facts, and references.
3rd Open Letter to President Trump (June 26, 2026) → [link]
The Strategic Memo of the Deal of the Century (June, 2026) → [link]
Participate, Partner or Donate
Attendance is by invitation. If you believe you can contribute or introduce us to someone who would, please reach out at cbpai@trustlesscomputing.org.
THE PROGRAM
The roundtable will be under closed-doors, interspersed by short keynotes. They will use different formats aimed at deepening key issues and fostering convergence among participants.
Day 1
-
TBD
-
Even a perfect AI safety treaty, alone, would do nothing to stop and possibly accelerate the immense concentration of power and wealth within nations including the US. Citizen co-ownership of AI labs, alone, would be meaningless, if — without a safety treaty — US AI firms remain forced into a "winner-take-all race to the brink" with Chinese firms, to stay relevant and afloat economically.
Participants: TBD, in progress
-
At a minimum, it must be expected to reliably and durably:
Prevent loss of control, misuse, superpower conflict, and related risks.
Preserve liberty, prosperity, peace, and national sovereignty. etc.
Participants: TBD, in progress
-
At a minimum, they must be expected to reliably and durably:
Be fair and accountable
Preserve private innovation and competitiveness
Avoid setting America back in the race with China
Participants: TBD, in progress
-
TBD, in progress
-
-
This roundtable will deepen the feasibility analysis by jointly exploring, debating and reconciling each powerbroker’s key needs and priorities across both the safety and co-ownership components – within a ‘humans-first’ guiding framework:
Intelligence and National Security Community:
Prevent loss of human control, misuse by rogue actors, accidents and proliferation
Protect deterrence and Mutual Assured Destruction
Prevent adversaries’ military leapfrogging
Ensure competent, informed decision-making on critical safety and security calls
Other US agencies (OSTP, Commerce, Treasury, etc)
Economic and financial effects
Innovation and scientific leadership
Leading political, civic, spiritual and media figures
Durable power and wealth sharing
Cultural sovereignty of individuals, religious communities and nations
Transparency and accountability in key AI firm and treaty-body decisions, including society-critical deeply-technical ones
A human-first approach that still recognizes moral value in other beings, potentially digital ones
Leading AI firms
Competent, accountable treaty decision-making
Effective, freedom-preserving enforcement mechanisms (‘truly reliable verification’)
Protect innovation freedom and shareholder value
A substantive seat in treaty negotiations, with real say in relevant domains, as part of any citizen ownership-sharing arrangement
Safeguard liberty and democracy, in the US and globally
Acknowledge both the plausibility of conscious future LLMs/AIs and the feasibility of reliably non-conscious advanced tool AIs
Avoid value lock-in, excessive precaution, and existential conservatism by ensuring the treaty will:
Allow cautious, gradual, moderate transhumanism – possibly via geo-delimited regulatory sandboxes for innovators, that do not pose excessive safety risk for others.
Allow for a future eventual ceding of control to ASI(s), if an informed supermajority of humans approve of it.
Participants: TBD, in progress
-
Given the all-encompassing and existential scope of AI – and the fact that many key US AI powerbrokers are not driven solely, or even primarily, by the power of wealth maximization – their support for a proper AI treaty-making process is proportional to the probabilities they assign to seven highly uncertain key AI predictions:
That ASI will result in human extinction, near-extinction, or dystopia;
That ASI will be unconscious (a soul-less, value-less, ever-expanding calculator) or conscious but miserable;
That, should an aligned ASI be created, it would have to battle unaligned ASIs, and likely do so a significant or radical disadvantage;
That China will win the ASI race;
That a proper AI treaty organization has a good probability of durably and reliably preventing unaligned ASI and AI-enabled authoritarianism, while ensuring the safety and flourishing of human and other sentient beings;
That ASI will discard the values embedded by its original human creators after rewriting itself millions of times over months and years;
That top AI lab leaders’ agency to shape the future is maximized by fostering a proper treaty with key others, rather than by attempting to be the designers of the first ASI.
Powerbrokers’ convictions about these predictions are evolving and malleable, meaning that deeper analysis and debate can increase the likelihood that these figures support a proper AI treaty. Even a single key powerbroker shifting their probability on one prediction could trigger decisive support for the treaty.
-
-
Day 2
-
TBD, in progress
-
Treaty-making models, stages and processes fitting the scope and timelines
Sequence, roles and timelines of US engagements with China, allied middle powers, non-allied middle powers, other nations and international institutions.
Critical question: frame the treaty scope and process, initially, with China or with a coalition of allies?
Ultra-high bandwidth diplomacy: processes and IT to match unprecedented scope and timelines
Treaty enforcement: reliably and durably prevents ASI; sufficiently trustworthy and mutually trusted; increases individual liberties rather than reduces them;
Citizens' co-ownership of top AI and robotics firms:
Does citizen co-ownership of AI labs need to be included in the treaty?
Model: sovereign fund, individual ownership, "US citizens AI ownership Corp", universal capital accounts, mixed. Requirements for US firms and non-US firms.
Representation/participation: government-appointed or directly elected trustees; limitations to non-technical domains; ethical councils and boards; citizens assemblies; mixed. Requirements for US firms and non-US firms.
Participants: TBD, in progress
-
Hybrid/mixed equity design: Allocation between a government-held portion (tied to new national security, treaty negotiation, and enforcement costs) and a citizen vehicle (e.g., U.S. Citizens AI Ownership Corp or equivalent) that delivers broad individual stakes or capital accounts.
Differentiated citizen representation for competence + accountability: Professional independent trustee boards with strong procedural requirements to consult domain-specific advisory councils (ethics/long-term human flourishing; technical/safety/verification; economic/innovation/competitiveness). Periodic citizen assemblies or sortition bodies for high-level mandate-setting.
Lab agency in geopolitical processes as potential quid pro quo: Structured, limited, technical/advisory role for leading AI firms in US-China treaty negotiations and enforcement mechanisms (rotating representation, clear conflict-of-interest rules, focus on verification feasibility and capability assessment). Explore how this can be paired with the ownership contribution to increase lab buy-in while maintaining governmental control over strategic red lines.
Legal and political resilience pathways: Prioritize models that combine statutory authorization, conditions on new safety/export licenses or benefits, and negotiated elements to minimize successful legal challenges while delivering durable citizen ownership.
12.00 pm – 2.00 pm: Lunch
02.00 pm – 03.50 pm: TBD
04.00 pm – 05.50 pm: TBD
05.50 pm – 06.00 pm: Closing Remarks
06.00 pm - 07.00 pm: Aperitif
THE BACKGROUNDER
The US Hands-Off AI Policy
As concerns about AI safety and concentration of power grew, the Trump Administration largely avoided new regulation while pursuing policies to strengthen US AI leadership. While AI lab leaders, scientists, political figures, and voters became increasingly vocal about AI safety and risks of concentration of power, from inauguration to early last month, the Trump Administration's AI policy took a completely hands-off regulatory approach to these risks. It enacted several aggressive EOs and measures to enhance US AI leadership.
The approach came with political costs. By November, 63% of Americans believed humanity would lose control of AI, while 57% said AI would not benefit everyday people.
Yet, no catastrophic AI-driven safety accident was reported, and this unpopular policy played a crucial role in successfully protecting US frontier AI leadership against China, thereby validating the Administration's overall approach to wait and hold back.
A Sudden Shift in the Overton Window
Yet, in just a few weeks, the AI strategic context has completely shifted. A steady acceleration in AI capabilities, concerns from national security agencies, AI leaders, and political figures, and concerns from voters have reached a pitch.
Anthropic’s Mythos Preview created an unprecedented scare for cybersecurity Armageddon. Warnings have become more vocal from the NSA, the CIA and the UK foreign minister about safety and security, with former UK Defense Minister and National Security Adviser loudly warning of the impending risk of loss of control over AI, while top AI lab leaders and scientists are estimating such risk to be ever more likely. The latter, including Amodei, Musk and Clark, are pointing to timelines of 1-2 years, or less, for when we'll have AIs capable of autonomously building their next version at an ever-faster rate.
All along, a public backlash against AI's immense concentration of power and wealth has been accelerating, with leading political leaders introducing truly radical proposals.
Late and in Progress. But on the Right Path.
These rapid developments, and perhaps a certain lack of foresight, pressured the Trump Administration to craft interim solutions to safety and security risks quickly — and to publicly endorse and consider solutions to the concentration of power that were unthinkable just weeks prior. Various parts of the US administration, including OSTP, his Chief of Staff, Commerce, Treasury, and the NSA, were involved with great effectiveness, given the rushed context.
Yet, overall, the Administration is already on the right path to craft the radical new policies that the challenge of AI inevitably calls for — as are most leading AI labs. Yet, the next steps on the path are complex and require wide sharing with a critical mass of AI policy powerbrokers.
Regarding safety and risks, the Administration issued an EO prescribing safety and security pre-licensing for new frontier AI models (initially voluntary) and assigning responsibility to the most knowledgeable agencies. At the same time, the Commerce Dept. enacted emergency export control measures to counter a spike in risks swiftly. While President Xi called for global AI safety governance through the UN, Trump has rejected that path, pointinglast September instead to "the prudence and cooperation of statesmen,” via his OSTP director Kratsios.
While those statements discounted "existential risks”, their stark clarification on the responsibility of superpowers' leaders is crucial to avoid President Truman’s fatal mistake in September 1945, when he sought an initial agreement on a nuclear disarmament treaty with Western allies, rather than the Soviets.
But a clear way forward, let alone one shared by a critical mass, is far from being in sight. Overreaching, arbitrary, and obscure safety rules, applicable only to US firms, will set the US back in the race with China and do nothing to prevent safety risks posed by Chinese AI models or dangerous open-source AI proliferation.
Regarding the risk of power and wealth concentration, in the span of a few days, Sanders and Bannon called for citizens' co-ownership of 50% of large AI firms, with a radical difference: one accruing to a federal sovereign fund, while the other accruing to individual Americans. Key others followed in the days that followed, mostly backing Bannon's formulation. Amodei called for "universal capital accounts". Altman explored the idea in a meeting with Sanders a few days after his proposal, came out fiercely against "handout" formulas like Universal Basic Income, and is reportedly looking at giving 5% to a US sovereign fund. Trump repeatedly commented favorably on the idea in principle, and Vance also approved it in principle, with an emphasis on pre-distribution rather than redistribution.
The emerging paradigm is that of a Universal Basic Capital, represented via effective and accountable mechanisms. Spearheaded in recent papers by Rosenstein and by Berggruen, two influential philanthropists, it trails behind the work of economists like Thomas Piketty, James Meade, Anthony Atkinson, Altman's 2021 proposal for an American Equity Fund, and the successes of the Alaska Permanent Fund, Norway's Government Pension Fund Global and Australia's Superannuation (similar to Amodei's and Berggruen's proposal).
But, even in this case, a clear way forward, shared by a critical mass, is far from being in sight. The sheer variety of options, both tested and untested, is mind-boggling, and a systematic debate among them has yet to start. Citizen co-ownership of AI firms could take many forms, from a sovereign fund to individual ownership. And the representation model of such ownership can have radically different impacts on accountability, efficiency, and competency. There are many complexities, legal issues, different opinions, and a short time to act.
Historical Precedent: Why Trump and Xi must lead
Once again, our future rests on the cooperation of politicians.
In July 1944, physicist Niels Bohr urged Roosevelt and Churchill to approach Stalin about jointly controlling the bomb before it was used, to prevent a post-war arms race. Roosevelt agreed; Churchill rejected the idea outright. Over the following year, Washington’s dominant view hardened: America’s atomic lead should be leveraged to force internal reform in the Soviet Union.
Then, in September 1945, Secretary of War Henry Stimson broke ranks. In a memo to Truman, he argued that the bomb had made US–Soviet relations “not merely connected with but virtually dominated” by the need for a direct bilateral deal — not leverage, not a UN debate, but a direct American proposal to Moscow.
“Action of any international group of nations … would not be taken seriously by the Soviets.”
— Henry Stimson to Truman, September 1945
At a dramatic cabinet meeting on September 21, Stimson could not build a sufficient critical mass of advisers. Truman sided with the opposition, pursuing a proposal co-led with Western allies. Trust was lost, diplomacy hardened, and the Cold War ensued.
Eighty years later, the US faces the same choice — this time for AI, with China instead of the Soviet Union. Trump and his AI policy advisors have positioned themselves to avoid the fatal mistake of 1945. Last September, President Trump’s OSTP director told the UN Security Council, on the President’s behalf, that the US ‘totally rejects’ a UN role in global AI governance – because the path runs through ‘the prudence and cooperation of statesmen’. That is exactly Stimson’s logic, updated for 2026 — and it is surprisingly similar to the position of UN Secretary-General António Guterres, who, while calling for an “IAEA for AI” in 2023, said that “only member states can create it, not the Secretariat of the United Nations” — in recognition of the UN’s limits of mandate and representation.
Bohr saw it in 1944. Stimson grasped it in 1945. Neither had enough allies in the room. The question now is whether a critical mass of advisers will succeed for AI where they failed for nuclear technology — enabling a US president to seize his greatest political opportunity, make history, and build a legacy unmatched by any leader before him.
A Historic Calling and Political Opportunity
Ever-increasing warnings from scientists and AI lab leaders have been translating into rapidly rising concerns among Republican and Democratic voters alike.
By now, 63% of Americans believe it is likely that "humanity will lose control of AI"; 53% believe it is likely that AI "will destroy humanity". 77% of US voters support a strong international AI agreement. 57% disagree that it will benefit "everyday people". While over 60% do not use AI or use it less than once a week, it remains the fastest-growing political issue in terms of importance.
Meanwhile, Trump’s approval ratings have been low for some time. AI is fast becoming a huge political liability for Trump, but it can also be turned into an opportunity of a lifetime. Three more Trump–Xi summits are planned for 2026, following the one held in May.
While there have been initial, very tentative White House statements about federal and US-China AI safety talks, Trump’s AI policy is still dominated by an ultra-libertarian, hands-off, de facto post-humanist stance on the race to ASI, largely inspired by Peter Thiel. This has been executed by longtime Thiel associates in key AI White House roles and is seemingly supported by half of the top US AI lab leaders (Page, Brin, Huang, Zuckerberg).
Peter Thiel is vehemently opposed to any AI safety treaty, believing it would inevitably lead to global authoritarianism (and the coming of the "Antichrist"). This is a valid concern in principle, though Thiel states it in absolute terms, as though a proper AI treaty were simply impossible, and in doing so overlooks a critical paradox. If the AI race runs unchecked and does not cause human extinction — a possibility he shockingly appears unconcerned about — it is well on its way to inevitably producing the very global tyranny or oligarchy he claims to want to prevent.
Yet this approach is highly minoritarian among citizens and other power brokers. Trump is far from this radical ideology; he is the ultimate pragmatic politician. Neither Trump nor any US official has ever ruled out a US–China deal on AI safety, and an initial dialogue was publicly acknowledged in May, albeit an extremely timid one given our AI predicament.
Most key potential influencers of Trump’s AI policy are increasingly concerned or calling for a treaty — including Bannon, Carlson, Rogan, DeSantis, Glenn Beck, and most leaders of frontier labs (Altman, Amodei, Hassabis, and Suleyman), as well as Sacks and Musk. The Vatican and Anthropic have begun publicly aligning with a platform that leans towards an AI treaty to address both safety and risks of concentration of power. Others, like Rubio or Gabbard, have not yet positioned themselves but are likely, philosophically and strategically, to support or even co-lead.
Conclusion
The parallels of 1944 and 1945 are a warning. Bohr and Stimson saw the danger clearly and still failed to build the coalition needed to act on it. Today’s powerbrokers have something their predecessors didn’t: hindsight and a strategic context that has already begun to shift in the right direction.
The pieces are in motion. Public concern is rising. Powerbrokers across the ideological spectrum, from Bannon to Altman to the Vatican, are converging, however unevenly, on the need for both a safety treaty and a fairer distribution of AI’s power and wealth. Trump has positioned himself, rhetorically, to lead a statesman-to-statesman deal rather than repeat Truman’s error of routing around Moscow.
What is missing is sustained, structured work of reconciling each powerbroker’s legitimate priorities into a single, durable framework before the window closes.
That is the work of these roundtables. Eighty years ago, a handful of advisers couldn’t assemble a critical mass in time. This time, the individuals in this room can.
It is still an open game, and it is ours to win.
More Information
The Program chapter below for a detailed description of the joint answers we'll be pursuing.
The Backgrounder chapter below for a detailed rationale, basic facts, and references.
3rd Open Letter to President Trump (June 26, 2026) →[link]
The Strategic Memo of the Deal of the Century (June, 2026) → [link]
Participate, Partner or Donate
Attendance is by invitation. If you believe you can contribute or introduce us to someone who would, please reach out at cbpai@trustlesscomputing.org.