Discussion paper  ·  Grey Swan / Archimedes  ·  June 2026

A 2050 AI Settlement

Grey Swan is mainly a warning and likelihood framework. This paper asks the complementary question: what would a genuinely better 2050 require if technological capability is to become shared human capability?

It is deliberately not a scientific prediction of 2050 and not a single policy blueprint. It is a positive systems proposition: the major systems around AI, climate and political economy have to move together, and the gains have to reach beyond frontier firms, capital owners and already-advantaged states.

June 28, 2026
Date
Discussion paper
Format
v11.9
Framework basis
Spring 2026
Evidence basis
Model overview Spring 2026 Fall 2026 2050 Settlement
Why this paper sits here

From diagnosis to a positive agenda

The scenario framework identifies the risks. The settlement paper asks what a credible positive trajectory would actually have to contain.

The question

The paper starts from a simple problem. AI may raise productive capability dramatically, but aggregate capability is not the same thing as broad social benefit. A better outcome depends on who controls capability, who carries transition risk, who owns the upside and whether public systems can keep pace.

The paper therefore shifts the emphasis from predicting AI capability to designing the conditions under which capability can be translated into security, agency and participation.

The proposition

A positive 2050 is not AI plus clean energy. It requires resilient infrastructure, health, education, trusted institutions, workable international coordination, meaningful productive access to AI and economic arrangements that spread gains.

The paper is global rather than country-specific and intentionally plural about policy instruments. Its claim is about system dependencies, not one predetermined political program.


The settlement at a glance

Six pillars, climate underneath them, wealth diffusion as the test

No single commitment is sufficient on its own.

01

Energy and grids

Clean, reliable and affordable power as social infrastructure, not merely announced generation capacity.

02

Health security and care

Prevention, early warning, distributed care and enough workforce capacity to remain resilient under stress.

03

Education and lifelong capability

Agency in an AI economy rather than dependency: people able to learn, judge, adapt and move between roles.

04

Trust and information integrity

Enough verification, transparency, audit and appeal capacity for public life and high-impact systems to remain governable.

05

Treaty-lite coordination

Practical, bounded cooperation around shared risks, supply security and avoidable cross-border breakdowns.

06

Compute and AI access

Productive capability beyond subscription access, including universities, SMEs, public systems and non-frontier countries.

Underlying condition

Climate mitigation and adaptation

Climate is the physical stress condition under which every promise in the settlement either holds or fails.

Decisive test

The Wealth-Diffusion Gate

Productivity, resilience and capability only count as empowerment when gains reach households, workers, SMEs, regions and public systems in observable and persistent ways.


Why distribution is binding

Growth is not the same thing as empowerment

The settlement refuses to assume that productivity gains will diffuse automatically.

The gate exists to make the claim testable.
If AI and clean energy increase productivity while the benefits accumulate mainly to frontier firms, capital owners and already-advantaged states, the positive scenario has not been reached. The paper therefore treats ownership, public returns, employee participation, portable benefits, stronger services, SME capability and transition security as part of the settlement rather than as an afterthought.

Consumer access is not enough

Millions of people can have access to capable AI services while firms, universities and public institutions remain dependent on external providers for compute, data infrastructure, capital and expertise. The settlement therefore distinguishes access to a product from the ability to deploy AI productively.

Capability is not only technical

Education, institutions and public systems matter because technology only becomes socially useful when people and organizations can use it with agency, judgment and resilience. The settlement treats these systems as part of the productive architecture itself.


A useful caveat

A positive systems proposition, not a solved plan

The paper is deliberately ambitious, but it does not pretend the politics have been solved.

The settlement identifies leverage points and makes their dependencies visible. It does not claim that one package can be specified in advance for every country, or that good policy ideas implement themselves. Some levers are easier to describe than to deliver, and the political economy of ownership, bargaining power, taxation, public returns and institutional reform remains the hard part. The value of the paper is to make the standard for a genuinely positive outcome explicit rather than assuming that technological progress will deliver it automatically.


Read the paper

A 2050 AI Settlement

Simpson, E. (2026). A 2050 AI Settlement. Zenodo. https://doi.org/10.5281/zenodo.21292104

Open access

Read or download from Zenodo

The Zenodo record provides the full discussion paper and a permanent DOI. The paper was written against the Spring 2026 v11.9 evidence base; the Fall 2026 v12 update changes the measurement architecture but not the paper's role as the positive companion to Grey Swan.