Google DeepMindP282026-07-06full textai ethicsdistributive justicetechnology policyhuman rightsglobal inequality

The Case for Globally Beneficial Technology

Gabriel and Kasirzadeh give five separate moral arguments that the material benefits of advanced technology, including AI, morally belong to everyone in the world, not just to inventors or firms.

It is DeepMind researchers' own published ethical case for why AI's economic benefits should be distributed globally, which bears on how the lab frames its mission and product/access decisions.

Iason Gabriel · Atoosa Kasirzadeh — meet the researchers →

How much of this do you want?
Orient me keeps four things: the abstract, the method, the claim↔evidence panel, and where it leads next. Everything adds constructs, the model table, every reported statistic, the discussion framing and the style moves. Switching hides nothing permanently and never changes what a section says — it only changes how many are on screen.
Abstract

Two readings, equal authority

How to choose: The paper’s words is verbatim — use it when you need to quote, or to judge how they write. Plain language is a paraphrase written for comprehension — use it when you want the idea fast. Neither is a summary of the other; they are two doors into the same room.

“To whom do the fruits of advanced technological innovation belong? To their inventors, to the organizations and individuals involved in making such discoveries possible, or to still larger groups of people, potentially encompassing all of humanity? This question sits at the heart of the present investigation. The arguments developed here focus on an expansive reading of the entitlement to benefit from technological breakthroughs: we argue that they should be designed, developed, and distributed in ways that benefit everyone. This central claim, which encompasses technologies such as advanced forms of artificial intelligence, is grounded in an exploration of five moral arguments that involve human rights, beneficence, contingencies of birth, the global tree of knowledge, and global economic justice. Taken together, they underpin the argument for globally beneficial technologies.”

Constructs

What this paper defines

Every definition below is the paper’s own sentence, with its locator. The plain gloss is a reading aid and is marked as one.

The Central Claim

“the benefits of advanced technology belong to the world. More precisely, we provide five arguments in support of what we term the Central Claim, which holds that advanced technologies belong to the world in the sense that everyone has a claim to benefit materially from their invention and use.”Introduction

In plain terms: Everyone in the world is morally entitled to a material share of the benefits created by advanced technology, not just its inventors or owners.

Transformational (technology)

“A technology is transformational if it has a profound effect on many aspects of life.”Introduction

In plain terms: A technology counts as 'transformational' if it substantially changes many parts of life.

Beneficial (technology)

“A technology is beneficial if it can be expected to create enormous positive value, conditional upon the way in which it is distributed and used.”Introduction

In plain terms: A technology counts as 'beneficial' if it can create huge positive value, depending on how it is shared and used.

General purpose technologies (GPTs)

“they are pervasive and perform a function that can often be adapted to different tasks; they are dynamic insofar as they make possible temporarily extended sequences of gains; and they spur complementary innovations that extend the value of the initial invention and lead to further use.”Introduction

In plain terms: GPTs are widely-usable, adaptable technologies (like electricity or AI) that keep unlocking gains and spark further inventions built on top of them.

Technologists

“We use the term "technologists" to denote the individual and institutional actors who exercise decision-making authority over the design, development, funding, management or deployment of a technology.”Section 1

In plain terms: 'Technologists' means anyone with real decision-making power over a technology's design, funding, or rollout, not just engineers.

Global tree of knowledge

“Modern science and technology rest on the global tree of knowledge, an immense intergenerational stock of ideas, theories, data, and methods accumulated over millennia”Section 4

In plain terms: The whole accumulated body of human scientific and technical knowledge built up across generations, which every new invention draws on.

Communalism (norm of science)

“the norm of communalism (Merton, 1973), which dictates that people who draw out of the pool of knowledge should in turn contribute to it, thereby supporting the wider cascade of beneficial innovation.”Section 4

In plain terms: The scientific norm that if you draw on the shared pool of knowledge, you owe something back to it.

Public goods (non-rivalrous / non-excludable)

“Public goods have a number of important properties. They are non-rivalrous in the sense that if one person uses it, this does not deplete the stock available for others (Samuelson, 1954). They are also non-excludable in the sense that it is hard, or impossible, to prevent other people from benefiting from these goods once they exist”Section 4

In plain terms: A public good is something one person using it doesn't use it up, and it's hard to stop other people from benefiting from it too.

Method

What they actually did

Each step is a synthesis. Open any step to see the paper’s own sentence it was derived from, with its locator — so nothing here floats free of the source.

Five independent moral arguments, each developed in its own section, converge to support the paper's single Central Claim rather than any one argument carrying it alone.
Click any box to open it.
  1. Opens with two historical case studies (concealed obstetric forceps vs. openly-patented seat belts) to make the stakes of how technology benefits are distributed concrete.
    Trace this step to the paper
    “These two inventions show that the way in which the benefits of technological innovation are distributed can be decidedly important for the world and for those who live in it.”Introduction
  2. States the paper's central thesis, the 'Central Claim,' that advanced technology's benefits belong to everyone materially.
    Trace this step to the paper
    “the benefits of advanced technology belong to the world. More precisely, we provide five arguments in support of what we term the Central Claim, which holds that advanced technologies belong to the world in the sense that everyone has a claim to benefit materially from their invention and use.”Introduction
  3. Grounds the investigation in background facts about global material scarcity and inequality that any distributive claim must be evaluated against.
    Trace this step to the paper
    “material scarcity and extreme poverty continue to be a reality of life for many people worldwide. One in ten people live on the equivalent of less than $3 per day, and over a billion people currently lack access to essential goods and services (World Bank, 2025).”Introduction
  4. Lays out a roadmap of five sections, each developing one independent moral argument for the Central Claim.
    Trace this step to the paper
    “Section 1 looks at the relationships between technology, human rights, and essential goods.…Section 4 looks at claims to benefit from technology arising from the “global tree of knowledge,”…Section 5 looks at the relevance of global economic institutions and background injustice for the Central Claim.”Introduction
  5. For each argument in turn, presents its case, then explicitly raises the strongest objections against it before responding to them.
    Trace this step to the paper
    “Nonetheless, when viewed as a potential grounding for the Central Claim, the Human Rights and Essential Goods argument encounters three challenges.”Section 1
  6. Synthesizes the five arguments in the closing section, arguing they converge on the Central Claim despite resting on different premises.
    Trace this step to the paper
    “This paper has explored five moral arguments in support of the Central Claim: the proposition that the benefits of advanced technology belong to the world in the sense that everyone is entitled to benefit materially from their invention and use.…they achieve a striking convergence—forming an overlapping consensus in support of the view that the benefits of advanced technology should accrue to all.”Section 6
  7. Derives practical implications, framing the Central Claim as agnostic about which institutional mechanism delivers benefit, rather than prescribing one.
    Trace this step to the paper
    “The Central Claim itself is largely mechanism agnostic. What it provides is a set of desiderata against which institutional proposals and actions ought to be assessed.”Section 6
The models under study

Exactly what was run, and how

What they reported — and what they left out

This is a normative philosophy paper, not an empirical study of AI systems: it discusses AI and other advanced technologies in general terms throughout and never names a specific model, version, temperature, reasoning-effort level, or deployment setting.

Results

The numbers they report

The paper cites the seat belt as a case where an openly-shared invention produced a very large estimated reduction in deaths.

an estimated 250,000 to 350,000 lives saved in the United States alone (Kahane, 2015; Sauber-Schatz, 2016)

See it in the paper
“Reasonable current estimates suggest that the modern seat-belt design has saved somewhere in the region of 250,000 to 350,000 lives in the United States alone (Kahane, 2015; Sauber-Schatz, 2016).”Introduction

A large share of the world's population currently lives in severe material scarcity, which the paper uses to motivate its distributive argument.

one in ten people live on less than $3/day; over one billion lack access to essential goods and services (World Bank, 2025)

See it in the paper
“One in ten people live on the equivalent of less than $3 per day, and over a billion people currently lack access to essential goods and services (World Bank, 2025).”Introduction

Access to digital technology itself is highly unequal globally.

two billion people currently lack Internet access (ITU, 2025)

See it in the paper
“with two billion people currently lacking Internet access (ITU, 2025).”Introduction

Place of birth has a dramatic effect on child survival odds, cited as evidence for the unfairness of unchosen circumstances.

a child born in sub-Saharan Africa is fifteen times more likely to die before age five than one born in a high-income country (UNICEF, 2025)

See it in the paper
“a person born in sub-Saharan Africa is fifteen times more likely to die before their fifth birthday than someone born in a high-income country (UNICEF, 2025).”Section 3

Hunger and child stunting remain widespread under the current global economic system, cited as evidence that existing global institutions are unjust.

hunger and stunting affect 650 million people and 150 million children, respectively

See it in the paper
“high levels of global material inequality, and the prevalence of hunger and stunting—which affect 650 million people and 150 million children, respectively—provide strong prima facie grounds for thinking existing global institutions are unjust.”Section 5
Claim ↔ evidence

What they assert, beside what they showed

Left is the claim in the paper’s own words. Right is the data offered for it. Where the two do not fully meet, a gold band names the distance.

The claim

The Human Rights and Essential Goods argument gives real but limited support for the Central Claim: it only reaches until basic needs are met.

“The Human Rights and Essential Goods argument therefore provides strong but conditional support for the Central Claim.”

The evidence

“while the Human Rights and Essential Goods argument provides strong support for the transfer of technology or technological wealth in situations where there is significant unmet material need, this claim seems to lose its force beyond a certain point—and to run out altogether once underlying needs are met.”

Section 1
Mind the gap: This argument alone cannot ground the fully universal, open-ended Central Claim stated in the Introduction, since the authors concede it 'runs out' once needs are met — which is exactly why four further arguments are introduced.
The claim

Even a weaker, minimal-cost account of beneficence still obligates wide distribution of an advanced technology's benefits.

“by distributing the benefits of an advanced technology widely, those who develop it are often in a position to do something very significant for humanity at little or minor cost, hence it would be wrong not to do so.”

The evidence

“the most morally relevant conception of “cost” understands this notion in relation to the potential gains (Ashford, 2003) that can be unlocked by an action. On this view, it seems likely that relative to the potentially vast global gains enabled by the wide distribution of benefits from advanced technologies,…most of the factors discussed above remain relatively minor.”

Section 2
Mind the gap: The 'minor cost' claim depends on comparing cost to potential gains rather than measuring absolute cost, and the authors themselves note that other moral views hold the absolute cost incurred is always relevant, leaving the premise contested rather than settled.
The claim

The unfairness of unequal life chances by place of birth generates a responsibility to redistribute technology's benefits.

“the Global Circumstances of Birth argument holds that the responsibility to use technology in this way is a by-product of responsibility that we all have (either directly or indirectly) to try to give people a fair chance in life.”

The evidence

“a person born in sub-Saharan Africa is fifteen times more likely to die before their fifth birthday than someone born in a high-income country (UNICEF, 2025).”

Section 3
Mind the gap: The statistic establishes unequal life chances by birth, but the further step that this specifically obligates redistributing technology, rather than other remedies, is argued separately and is contested by those who doubt unchosen inequality alone generates enforceable duties.
The claim

Because advanced technology draws on a shared, publicly-funded knowledge commons, technologists incur a reciprocal duty to distribute benefits broadly.

“Under these conditions, where commercial value is actively generated from the proceeds of a joint undertaking, the Tree of Knowledge argument holds that there is a responsibility to ensure broad material benefit.”

The evidence

“There is a huge publicly funded infrastructure—including the interlocking efforts of many people and institutions—that supports the continued generation, open distribution, and accessibility of knowledge at a global scale.”

Section 4
The claim

The current global economic system is unjust, which grounds a duty for those with power over technology to help remedy that injustice.

“The Global Economic Institutions argument supports the Central Claim because, in situations where background injustice exists, duties to remedy the situation are often widely distributed, extending to include all who are part of that system.”

The evidence

“high levels of global material inequality, and the prevalence of hunger and stunting—which affect 650 million people and 150 million children, respectively—provide strong prima facie grounds for thinking existing global institutions are unjust.”

Section 5
The claim

Taken together, the five arguments establish a robust (though not conclusive) case for a universal entitlement to benefit from most transformative technologies.

“We have established a robust pro tanto case for a universal entitlement to benefit from advanced technology, across a wide spectrum of cases, encompassing the majority of transformational GPTs.”

The evidence

“Ultimately, we find considerable overlap among the arguments, each of which supports the Central Claim in different ways. Some of the arguments considered here are powerful but not fully universal in scope. Others are universal but apply only to certain kinds of technologies or under certain conditions.”

Introduction
Mind the gap: The authors explicitly label the case 'pro tanto' (a reason that can be outweighed), and no single one of the five arguments is both fully universal and unconditional — the strength of the overall case rests on convergence across arguments rather than any one being individually decisive.
Discussion & after

How they frame it, and what they want next

Their framing

The authors frame their case as cumulative and pluralistic rather than resting on one decisive proof, explicitly rejecting a single knock-down argument in favor of several partially-overlapping moral arguments that together make the Central Claim hard to escape. They repeatedly pre-empt objections to their own position before answering them, and close by stating plainly which competing view they reject.

Register: The paper states its normative conclusion with high confidence, calling it 'a radical claim,' while explicitly hedging the completeness of its own case with terms like 'pro tanto' and repeated acknowledgments of what space limitations prevented it from covering.

Where they hedge

“We have established a robust pro tanto case for a universal entitlement to benefit from advanced technology”Section 6
“due to limitations of space, we are not able to evaluate all the arguments that bear upon the truth or falsity of the Central Claim.”Introduction

What they say it means

  • Institutional designs for distributing AI's benefits can be judged against the Central Claim as a normative standard, and found wanting.
    the paper’s words
    “Institutional arrangements that fail to account for these claims are presumptively unjust.”Section 6
  • There is no single correct mechanism for delivering technology's benefits; the right approach is empirical and depends on the specific technology and context.
    the paper’s words
    “the question of how to achieve widespread benefit from technology is empirical, and it needs to be answered differently for different technologies, in a way that is sensitive to their value and to any risk that they create.”Section 6

What they call for next

  • For transformative technologies like advanced AI, a mix of approaches is needed: open publication and model release, free-at-point-of-use market access, state provision, and coordinated global action.
    the paper’s words
    “technologies, such as advanced AI, it seems likely that a range of approaches are needed—including the open publication of research and release of models, market access that is free at the point of use, state-level service provision and coordinated global action to ensure that these benefits are unlocked and directed appropriately.”Introduction
  • Endorses building new multilateral partnerships to widen access to compute and talent, plus programs to make open AI models valuable across diverse local contexts.
    the paper’s words
    “a number of major studies have proposed the creation of new multilateral partnerships to widen access to computing power and talent (Dennis et al. 2025; United Nations, 2024), alongside programs designed to ensure that open AI models can be developed and refined in ways that render them appropriately valuable across diverse local contexts (UNDP 2025).”Section 6
  • Highlights the proposed 'Windfall Trust,' which would redirect a share of advanced AI's financial returns to all living people for globally beneficial outcomes.
    the paper’s words
    “More ambitiously still, a number of authors have proposed a “Windfall Trust” (O’Keefe et al., 2020)—envisaged as a trust for all living people, that would redirect a percentage of the returns from advanced AI towards globally beneficial outcomes.”Section 6

Limitations they state

“due to limitations of space, we are not able to evaluate all the arguments that bear upon the truth or falsity of the Central Claim. Nor are we able to provide a detailed roadmap for how universal benefit is to be unlocked in practice.”Introduction
“The arguments presented here are not intended to be exhaustive. There are further arguments, both in support and in opposition to the Central Claim, which have not been discussed—or were only touched upon briefly—due to limitations of space.”Section 6
“Without more comprehensive investigation of each argument and the way it bears upon specific advanced technologies such as AI, arriving at an all-things-considered judgment about what morality requires in any given case remains a complex task.”Section 6
For your own writing

Moves worth stealing

Opens with two contrasting historical case studies (a suppressed invention vs. an openly-shared one) to make an abstract distributive question concrete before stating the thesis.

“for more than three generations, the new instrument was kept a secret and used exclusively by the Chamberlens to bolster their business”

States the central thesis as a labeled, reusable term (the 'Central Claim') early on, then defines its scope precisely along named axes.

“First, it is fully universal insofar as it holds that everyone is morally entitled to benefit from advanced technology, not just certain individuals, organizations, nation-states, or some subset of the global population.…Second, the Central Claim is substantial insofar as it holds that everyone is entitled to benefit materially from this technology.”

For each of the five arguments, proactively raises the strongest objections against its own position before responding to them, rather than only presenting supporting evidence.

“Nonetheless, when viewed as a potential grounding for the Central Claim, the Human Rights and Essential Goods argument encounters three challenges.”

Closes by naming the specific position it rejects in one plain sentence, after already granting partial validity to nearby opposing views.

“What is not consistent with the Central Claim—and the position we reject in light of the arguments presented in this paper—is the moral claim to exclusive benefit from advanced technology.”

Discloses institutional affiliation transparently via an explicit disclaimer distancing the paper's personal argument from the employer's official policy.

“this paper represents the views of its authors and does not necessarily reflect the views or policies of Google DeepMind.”
Connected

Where else this leads

Same people

Published alongside it

The nearest publications in time, across all three labs.

What this page was built from

Working text is the full paper body, footnotes, and references as scraped from the source (corpus manifest grades it 'full'); this is a normative philosophy/policy paper rather than an empirical study, so schema fields built for empirical work (models_studied, methods as data collection, results as experiment outputs) are populated honestly by adapting them to the paper's argumentative structure and the background statistics it cites, rather than implying experiments that were never run.