---

Developers, governments, and those affected by artificial intelligence assign different significance to the same technical achievement.  A developer seeking commercial advantage may treat an advance as grounds for ownership.  A government concerned with strategic predominance may treat it as an instrument of national power.  Those exposed to its consequences may judge it through its effects upon employment, education, communication, or public judgment.  Each perspective depends upon relations that the achievement alone does not disclose.

These relations participate in development itself.  Investors influence which problems receive attention.  Owners determine who may use a capability and under what conditions.  Governments establish permissions, restrictions, and incentives.  The technology available to a user embodies decisions whose purposes need not be evident in its operation.

An answer appears on a screen.  Its fluency may lead the user to accept it as sufficient for understanding.  Yet the answer alone does not disclose how developers trained the system, which sources informed its response, what restrictions govern its output, or how developers organize its presentation to invite confidence.  The user may find the answer useful without knowing all these conditions.  That usefulness does not establish the adequacy of the answer.

Mediation names this dependence:  understanding rests upon relations that attention to its immediate object does not necessarily reveal.  The answer occupies the user’s attention.  The conditions supporting its production, presentation, and acceptance remain only partially visible.

The public accounts of artificial intelligence exhibit a comparable dependence.  Commercial accounts identify innovation through investment, demand, and return.  Strategic accounts identify technological capacity through advantage over competitors.  Security accounts identify uses through the dangers they introduce or prevent.  Each account directs attention toward relations relevant to its purpose.  Each leaves questions that its governing purpose cannot adequately answer.

These accounts also inform decisions.  An institution that measures achievement through commercial return has grounds to favor applications for which customers will pay.  A government that measures achievement through strategic advantage has grounds to favor capabilities that strengthen its position.  The criteria through which an institution recognizes value therefore bear upon the purposes it authorizes and the resources it assigns.  What remains outside those criteria may lack an effective claim upon its decisions.

Those who determine the terms of recognition can influence which needs count as reasons for action.  Their authority does not follow from the completeness of their account.  It may rest upon ownership, financial resources, statutory powers, or control over infrastructure.  An account can govern development and can still leave much of its social significance unexamined.

The United States’ 2025 AI Action Plan explicitly frames development as a race for global dominance, and links technological leadership with economic and military advantage.  The United Nations’ Global Digital Compact places international governance within a commitment to public benefit and inclusive participation.  The former sets out actions for federal agencies; the latter records commitments among participating states.  The purposes expressed in these documents operate through different institutional means.  Their coexistence supplies no common authority for deciding between them when their demands conflict.  ^[1]

Commercial success establishes that a product attracts expenditure.  It does not establish how its benefits and burdens are distributed.  National predominance establishes an advantage relative to other powers.  It does not establish that the advantage improves the circumstances of those subject to its consequences.  Protection against malicious use addresses a necessary concern.  Authorized uses may also diminish the ability of people to understand, question, or influence decisions affecting them.

The capacities under development nevertheless permit achievements of substantial human value.  Researchers can incorporate particular capabilities into successive stages of inquiry, and use them to examine complex relations, formulate hypotheses, and assist the design of inventions.  Particular applications may support advances in medicine, materials, or other fields concerned with human welfare.  The value of an achievement depends upon what it enables people to do and upon the consequences of its use.  Commercial ownership or strategic sponsorship does not exhaust that value.

The same capacity to assist inquiry introduces uncertainty when institutions delegate decisions and actions to systems.  Developers and users may permit a system to select means, initiate operations, or act upon other systems with limited intervention.  Its effects then depend upon the scope of that permission, the conditions governing its operation, and the ability of those responsible to inspect or interrupt its actions.

Agency in these circumstances need not resemble human intention to exert power over human affairs.  A system can influence conduct through recommendations that institutions routinely accept, classifications upon which access depends, or operations carried out under delegated authority.  Uncertainty concerns how its actions will interact with circumstances its designers and users cannot fully anticipate, and how effectively human beings can retain control over those interactions.

The system may also influence the relations through which people understand and authorize its use.  If an institution relies upon it to identify priorities, assess evidence, or evaluate performance, its outputs inform decisions about subsequent delegation.  Authority initially granted by human beings can therefore support operations that shape the grounds upon which further authority is granted.  The institution’s reliance upon the system extends beyond the performance of a task to the judgment of what tasks deserve performance and under whose direction.

Incremental benefit and uncertainty about agency can coexist within the same arrangement.  An invention may improve human welfare, and may increase reliance upon systems that those affected cannot readily examine or contest.  That reliance may rest upon cost, infrastructure, contractual restrictions, or the absence of practicable alternatives.  Limited understanding can impede scrutiny of the arrangement without constituting its sole cause.  The achievement retains its value.  That value supplies no sufficient account of the reliance established through its use.

The betterment of society consequently depends upon more than the intentions of those directing development.  If those institutions retain exclusive authority to define improvement, their criteria govern which consequences receive recognition.  A gain in productivity may accompany diminished security of employment.  Greater access to information may accompany greater reliance upon a small number of providers.  Expanded capabilities may leave those affected with fewer means of challenging their application.

Who receives the benefit, who bears the burden, and who possesses authority to decide are distinct questions.  An assessment confined to aggregate improvement can conceal the differences upon which its judgment rests.

The global reach of artificial intelligence extends those differences across societies.  Worldwide availability does not establish worldwide participation in determining its purposes.  A community may receive a system without the resources to adapt it to its language, examine its limitations, or challenge its terms of use.  Providers extend access while they retain control over the conditions of participation.  The technology serves a wider population without a corresponding distribution of authority over its development.

Nor does an appeal to the common good settle whose purposes should govern.  Judgments about the common good depend upon whose circumstances receive attention, which consequences carry weight, and what forms of reliance people are prepared to accept.  Institutions may affirm universal benefit while excluding affected communities from those judgments.

Technical competence, financial resources, political authority, security, and social need each sustain claims upon development.  None supplies a sufficient account of the whole.  Their unequal institutional power bears upon which claims govern, which remain negotiable, and which lack any effective means of expression.

Where institutions treat one criterion as sufficient, their judgments substitute a partial measure for the relations supporting it.  Commercial actors identify expansion with progress.  Governments identify predominance with protection.  Security authorities identify the prevention of prohibited uses with safety.  Providers identify availability with participation.  Each identification leaves dependencies outside the account through which the undertaking is justified.

The answer on the screen offers the user something immediately available for judgment.  Its usefulness may be real.  Its fluency does not disclose who determined its conditions, what authority supports its use, or how reliance upon it influences subsequent judgment.  The emerging capacities of artificial intelligence extend the consequences of those relations.  Those who delegate authority to systems require knowledge of the conditions governing the systems and powers to withhold, limit, or revoke that delegation.  Those affected without participating in the delegation stand in a different relation:  their ability to contest its consequences depends upon access to the grounds of decisions and institutional means of holding the delegating authorities answerable.

[1] The White House, America’s AI Action Plan, July 2025:  
https://www.whitehouse.gov/wp-content/uploads/2025/07/Americas-AI-Action-Plan.pdf

United Nations, Global Digital Compact, Annex I to the Pact for the Future, 2024:  
https://www.un.org/pact-for-the-future/en/annex-i-global-digital-compact

---