فارسی English
From Socialized Production and the Right to Decide to Project Meridian
Hamid Akhavi
Artificial intelligence is usually introduced through its capabilities: writing, coding, pattern recognition, and automating parts of decision-making. But to understand its social position, three questions must be asked together: who built this capability, who controls it, and how are the wealth and authority it generates distributed?
This article connects the interview guide and its supplementary proposals with the Meridian case study. At the center of the discussion is the gap that can open up between the socialization of production and the concentration of ownership and command. Progress must be measured by improvements in people’s lives: livelihood security, the quality of services, free time, and the ability to participate in the decisions that shape their destiny.
1. Ownership: The Right to Set the Rules of a Social Capability
Ownership is not just the deed to a building or a machine. It is a bundle of rights to determine access, price, terms of use, the direction of investment, and the distribution of returns. In AI, these rights are distributed across several layers: chips and computing power, data centers and the cloud, models and software, rights to use data, and the platforms that deliver services.
Millions may use a tool, yet have no effective role in changing its rules. Widespread use does not, by itself, create social ownership. When a system enters the workplace, the school, the hospital, or the information network, its owners’ decisions can affect the lives of people who are not even direct parties to the contract. It is here that ownership becomes a question of power: who decides, and who can challenge their decision?
Nor should all layers be treated as one. The openness of a model does not automatically eliminate the monopoly of cloud infrastructure; owning a data center does not necessarily mean owning all the data that passes through it. Precise analysis must show what authority exists in each layer and how the dependence of others is created.
2. Socialized Production and the Less-Visible Work
Chip design, model engineering, investment, and infrastructure management are real, costly work. The contribution of companies and specialists cannot be ignored. But this contribution rests on a far broader accumulation of knowledge, language, education, cultural works, research, and human labor. Some of these foundations were built with public resources, others through private, cooperative, or voluntary activity.
Human labor does not disappear behind the automated appearance of the system. Data preparation, evaluating responses, filtering harmful content, maintaining equipment, and producing energy are all links in this chain. A Time investigation of Kenyan workers at the contractor Sama reported that some text labelers for a project connected to OpenAI were paid about $1.32 to $2 per hour, and several of them spoke of the psychological effects of exposure to disturbing content. This report is a concrete example of working conditions, not a description of all workers in the industry. [1]
The question, then, is not whether to eliminate the share of the engineer or the investor; it is whether one kind of contribution justifies exclusive control over the outcome of the whole set of contributions. Recognizing the social origin of knowledge is also not a license for unauthorized use of personal information. Not all user conversations or clicks are necessarily used to train models; that depends on the service, the contract, and its settings. The right to privacy and consent must be preserved under every form of ownership.
3. Revenue, Profit, and Wealth: Three Different Concepts
Revenue is the money from selling goods and services; profit is what remains after costs and other accounting items are calculated; a shareholder’s wealth relates to the value of their assets. A rising share price can create paper wealth without that same amount ever being paid in cash. Share buybacks are not the same as a company’s operating profit or employees’ income.
Nvidia’s report for the second quarter of fiscal 2027 records about $96.2 billion in revenue and $59.7 billion in accounting net income. In the same period, the company returned about $26 billion to shareholders through share buybacks and cash dividends. Broadcom, for the third quarter of fiscal 2026, announced $16.7 billion in revenue from its AI-related semiconductor segment; that figure is the revenue of that segment, not the company’s net profit. [2, 3]
These figures show that large revenues and profits have formed in parts of the infrastructure chain. But on their own, they do not prove that all AI applications are profitable or that all profit stems from monopoly. The political question begins elsewhere: what needs are the resources obtained being spent on, and who decides that?
Workers may receive wages or shares, and pension funds may be shareholders; yet a financial stake does not necessarily create an effective right to decide. A fund’s assets under management should not be mistaken for the personal wealth of its manager. Critique of concentrated power grows stronger when these distinctions are made clear.
4. Private Risk, Social Cost, and the Bubble Question
Investment risk is real: the product may fail, infrastructure costs may never be recouped, and the company may lose money. But workers also bear the risk of job loss, income instability, and burnout. Society may contribute through education, the power grid, or research support, and may accept the environmental consequences. The amount and kind of these contributions must be examined for each project.
Rewarding investment is one matter; the right to set the rules of collective life is another. The existence of risk does not imply that any degree of wealth concentration or any kind of control over infrastructure is justified. Rights and obligations must be measured against the full set of contributions, supports, and consequences.
With data centers, the issue is not just the company’s electricity bill; the cost of new grid capacity, water consumption, pollution, and the prioritization of public services are also at stake. In the statement explaining his vote on the PJM matter, a member of the U.S. Federal Energy Regulatory Commission emphasized assigning capacity-shortage costs to regions with data-center growth. This source shows there is a dispute over cost allocation; it does not prove that all household bill increases in all regions are caused by data centers. [4]
At the same time, real revenue can coexist with optimistic valuations and a financial bubble. Chip sales are an economic reality, but share prices also rest on expectations about the future. Even if part of the market value collapses, the ownership question does not end: the infrastructure, the data, and the decision-making authority remain concentrated somewhere.
5. Open Models: Broader Access, but Not Necessarily Shared Power
Open models and software can expand the possibilities of inspection, modification, and innovation. Yet effective use requires skill, energy, and computing power. “Downloadable” does not always mean full freedom to use and modify; licenses and distribution terms matter.
A clear example is Nvidia’s announced agreement to acquire Hugging Face. The filing with the U.S. Securities and Exchange Commission cites about $11.9 billion in payments to shareholders and an employee retention equity program of up to about $1 billion. The deal is expected to close in the first half of 2027 and is subject to conditions and required approvals. The company has also committed to keeping the platform open and supporting other chip vendors. This agreement should therefore be called neither a completed purchase nor grounds for concluding that the platform will definitely be closed. [5]
Still, a commitment to open access is not the same as a distribution of governing authority. Users may be able to download the model yet have no say over development priorities, service pricing, or future rules. The defense of open access becomes more complete when accompanied by access to infrastructure and mechanisms of accountability.
6. Productivity for Whom: More Profit or More Free Time?
Suppose a task that took eight hours can now be done in six with the help of technology. Working hours could be reduced, quality improved, or more services offered. The number of workers could also be cut and more pressure placed on those who remain. Technology alone does not choose among these paths; ownership, contracts, organization, and political decision-making play their part.
Research by the International Labour Organization and NASK in 2025 shows that a quarter of the world’s workers are in jobs with some degree of exposure to generative AI. This estimate is not equivalent to the elimination of a quarter of jobs; the research considers the transformation of tasks more likely than full replacement. [6]
From this article’s perspective, rising productivity should make a better life possible: livelihood security, less exhausting work, and more time for care, learning, and participation. That outcome is not guaranteed. Workers must be able to negotiate the introduction of systems, monitoring of work, training, changes in tasks, and the distribution of productivity gains.
An example of turning this demand into enforceable rules is the 2025 SAG-AFTRA video game contract, which created protections around consent and disclosure for the use of digital replicas. This experience does not resolve the whole ownership question, but it shows that the terms on which technology is applied can be a subject of collective bargaining. [7]
7. From Owning Technology to Setting the Priorities of War: Meridian
This relationship takes on a more sensitive meaning in the military domain. When knowledge and computing capacity are applied to wartime decisions, the issue is not only the distribution of income; human lives and the risk of expanding violence are also at stake.
Pete Hegseth introduced Project “Meridian.” According to the official announcement republished in Forth, Elon Musk, Palmer Luckey, and Newt Gingrich share leadership of the study, with Emil Michael, the Pentagon’s chief technology officer, facilitating its launch. The mission is to examine the needs and capabilities of future wars; the expected output is a public report with a classified annex, on a 120-day timeline. [8]
The same announcement also separately announced the creation of an “Autonomous Warfare Command.” Meridian should not be conflated with this command, nor should the study’s leaders be presented as commanders of automated attacks. The documents reviewed also provide no basis for attributing any specific attack to this project.
The political significance of the group’s composition is clear: Musk brings space and communications infrastructure, Luckey brings defense technologies, and Gingrich brings a record of political leadership to this study. Bringing in specialists from outside government can provide useful experience; but expertise and commercial success are not, by themselves, representation of the public interest.
8. The State and Corporations: From Procurement Market to Influence over Defining Needs
The link between technology owners and military procurement can be shown through specific contracts. In the National Security Space Launch program, the U.S. Space Force announced an expected contract value for SpaceX of about $5.92 billion. This expected value should not be called realized revenue or company profit in a single stroke. [9]
The U.S. Army has also announced a framework for Anduril technologies with a ceiling of $20 billion, a five-year base period, and a five-year optional period. This framework covers software, hardware, and data and computing infrastructure, and is not simply a firm $20 billion counter-drone purchase. [10]
This evidence is enough to raise the question of conflict of interest: how does someone whose company supplies military technology take part in defining the future needs of the buyer? Are financial interests disclosed? Are recommendations independently evaluated? Who recuses themselves from related decisions?
This article’s structural analysis is that ownership of technical capacity can build advisory credibility; advisory credibility can affect budget priorities; and subsequent orders reinforce that same capacity and influence. This describes a potential mechanism, not proof of wrongdoing by Meridian. Judging the outcome requires examining the recommendations, the conflict-of-interest rules, and subsequent procurements.
In this relationship, the state is not merely a market observer; through research, procurement, and standard-setting, it plays a role in shaping the market. But a government contract is no guarantee of profit either. Precise critique must distinguish between financing, contract ceilings, actual payments, and profit, and instead of attributing definite motives to individuals, show the mechanism by which authority is distributed.
9. Military AI: Between Recommendation and the Authority to Use Force
Not all military applications of AI are the same. Image analysis, logistics support, target recommendation, and automated selection and attack involve different levels of authority. In announcing its contract with Special Operations Command, Anduril spoke of software for coordinating uncrewed systems and operators’ interaction with them. This is the company’s description of its product and contract, not independent confirmation of its safety or combat effectiveness. [11]
The danger is not limited to fully autonomous systems. The 972+ investigation of Lavender, based on the accounts of six Israeli intelligence officers, reported heavy reliance on the system’s output and very brief review of some targets. The Israeli military rejected this picture of automated targeting, emphasizing the role of analysts and independent review. This report and its response must be cited with clear attribution; the reporting sources’ claims should not be presented as undisputed fact or as a conclusion about all systems. [12]
The central issue is that nominal human presence does not guarantee effective control. Someone who lacks the time, information, or authority to reject a recommendation may become a mere rubber stamp. Responsibility cannot be offloaded onto the sentence “the algorithm decided so.”
The International Committee of the Red Cross has warned about weapons that select targets and apply force without human intervention. The institution recommends banning unpredictable systems and autonomous weapons that directly target humans, and strict limits on other types. These are the institution’s recommendations and should not be mistaken for an already-adopted global ban. [13]
The Anthropic–Pentagon dispute also shows that the limits of technology’s application are themselves contested. To describe the legal situation, the August ruling alone does not suffice: the D.C. Circuit Court of Appeals rejected the company’s challenge to its removal from the supply chain under the relevant statute. This path must be distinguished from the separate California litigation. Nor should this dispute lead to the conclusion that a single company represents democratic oversight of war. [14]
10. Redistribution of Income and Participation in Power
Taxes, public services, and income guarantees can strengthen the security of life. Someone constantly worried about food and rent has less opportunity and capacity to participate. Yet payments of money alone do not create a right to decide about data, the workplace, or the military application of technology.
Basic income must also be judged by how it is financed and its relationship to public services. If it replaces education, health care, and essential protections, it may reproduce part of the insecurity. The measure is increased real agency and improved lives.
State ownership is not, by itself, social ownership either. Transferring authority from a corporate board to an unaccountable state apparatus may preserve the concentration of command. Social ownership is meaningful only when people have a real role in setting goals, allocating resources, and evaluating outcomes, and can hold managers to account.
This participation can take different forms: accountable public infrastructure, worker cooperatives, joint university-public institutions, or combinations of these. Professional independence, contract transparency, minority rights, and privacy must be preserved. Not everyone needs to be an engineer to have a say about priorities in health care and education or resource use; specialists explain possibilities and consequences, but the legitimacy of setting public goals does not come from technical expertise alone.
11. From Critique to Measurable Change
The starting point is the rights and institutions that make participation practical. In the workplace, the right to negotiate over AI’s application must be accompanied by training, the ability to challenge algorithmic evaluation, and workers’ share of productivity gains. In the data domain, consent and privacy must remain enforceable even within public infrastructure.
Public support for companies must carry measurable commitments: public access, service quality, working conditions, and the return of benefit to society. Data-center projects must also make clear their water and energy costs, their impact on the grid, and their local commitments. Setting costs based on a project’s actual impact is more useful than generalizing from uncertain statistics.
In the military domain, this article proposes disclosure of advisers’ financial interests, rules for recusal from related decisions, independent review of recommendations, oversight of budgets by elected bodies, and clear limits on delegating lethal authority. Publishing the Meridian report could enable scrutiny, but accountability also requires the ability to influence and halt harmful decisions.
Public and cooperative computing capacity for universities, essential services, and small groups can reduce dependence — provided its management is accountable and its access rule-governed. No form of ownership is immune to mismanagement. Every plan must be judged by service quality, access, free time, working conditions, and environmental effects.
12. The Future of Technology Is a Question of Social Choice
Under capitalism, asset ownership and the expectation of profit shape the direction of investment. An application with higher financial returns may attract more resources, even if another social need is more urgent. AI raises this question on a new scale, because it can enter the production, management, and execution of decisions.
The widening gap between socialized production and private appropriation is this article’s political interpretation of this structure; it cannot be proven with a single profit report alone, nor does it imply that all companies act alike. Useful applications are real, just as the danger of concentrated power is real. Using existing tools does not take away people’s right to criticize ownership rules and demand their change.
Meridian carries this discussion from the question “who gets rich?” to the question “who defines the future?” A society that provides the knowledge, labor, and infrastructure should also have a share in choosing the direction of their use. AI can make it possible to reduce suffering and expand human capability; realizing that possibility depends on the distribution of income, the distribution of authority, and people’s organized capacity to change the rules.
References
[1] Time’s investigation of Kenyan workers and Sama [2] Nvidia’s financial report, second quarter of fiscal 2027 [3] Broadcom’s financial report, third quarter of fiscal 2026 [4] A FERC member’s explanation of capacity-cost allocation in PJM [5] Nvidia’s SEC filing on the Hugging Face acquisition agreement [6] The ILO and NASK research on occupational exposure [7] SAG-AFTRA’s explanation of AI protections [8] Forth’s republication of the Pentagon’s official announcement on Meridian [9] The Space Force announcement on National Security Space Launch contracts [10] The U.S. Army announcement on the Anduril contract framework [11] Anduril’s announcement on the mission autonomy contract [12] The original 972+ report on Lavender and the Israeli military’s response [13] The Red Cross’s explanation and recommendations on autonomous weapons [14] Text of the D.C. Circuit Court of Appeals ruling, republished on Justia
Editorial Note
This text was compiled by merging the interview guide “AI Ownership and Wealth,” its supplementary proposals, and the Meridian article; it is an analytical rewrite, not a transcription of a conducted interview. The selected examples and figures have been checked against linked sources. Additional unverified statistics and unnecessary biographical details from the preliminary texts were not included in the comprehensive version. Direct access to the Pentagon announcement was not possible, and the specified republication of the announcement in Forth was used instead.
Leave a comment