Welcome to the Machine

How Schemes to Control AI Are Progressive Tools to Control Society

The transition toward superintelligence will come with serious risks—from economic disruption, to misuse in areas like cybersecurity and biology, to the loss of alignment or control over increasingly powerful systems. Without effective mitigation, people will be harmed. Avoiding these outcomes requires building new institutions, technical safeguards, and governance frameworks so that advanced systems remain safe, controllable, and aligned.

—OpenAI, Industrial Policy for the Intelligence Age

AI industrial policy entanglement: Misalignment by design

AI governance proposals frequently invoke the goal of “alignment.” On its face, that’s just shorthand for ensuring that AI systems pursue the goals benevolent designers intend while minimizing the risks of unintended, harmful outcomes. But the megalithic core of many such schemes is something else entirely: the entanglement of business and government. This blurring of the private shareholder economy with federal cartelization and tight regulation will embed misalignment by design.

OpenAI’s ambitious new Industrial Policy for the Intelligence Age fits the pattern, painting AI as both a revolutionary opportunity and a looming existential risk requiring new institutions and governance frameworks that welcome vast new government powers.

The transition to AI could be managed without social upheaval if policymakers focused on practical reforms. Those include preempting over-regulation by the states and ensuring the power needs of data centers get met. However, many players see a benefit to themselves from promoting a perception of crisis. As The Economist pointed out, alarmism can even make commercial sense for dominant firms. A better policy approach properly identifies the entanglement of government and industry as the most imminent source of risk and misalignment.

The government-sized elephant in the room

OpenAI’s focus on private-sector AI harms glosses over the more consequential role government plays as an AI actor. As we have stressed with respect to emerging White House and congressional AI frameworks, the threat is less that private firms will misbehave than that government will. (These frameworks specifically are the White House’s National Policy Framework for Artificial Intelligence and Sen. Marsha Blackburn’s (R-TN) TRUMP AMERICA AI Act.)

While OpenAI’s plan commendably calls for guardrails to “establish clear rules for how governments can and cannot use AI,” it ultimately expands the federal AI footprint in ways incompatible with government accountability and the rule of law.

Federal agencies already deploy, fund, and in some cases indemnify or otherwise shield AI contractors, major AI firms, and adjacent technology sectors across communications, defense, energy, and health infrastructure. This federal presence already shapes incentives, risk profiles, and technical trajectories in ways no private-sector framework can counterbalance. In its claim that “market forces alone aren’t sufficient” to regulate AI, OpenAI fails to appreciate that market forces are preempted in the first place. Steering AI toward socially beneficial outcomes requires recognizing that political forces are what are insufficient, not market forces.

OpenAI claims the regulation of transformative technologies such as electricity, automobiles, and aviation provides a model for today. Those technologies, says OpenAI, “introduced new risks alongside new benefits, and new systems were built to manage them as they scaled.” In reality, such regulatory systems failed to scale effectively. They introduced rigidity where adaptability was needed. They kept key sectors artificially isolated from one another. These include communications regulated by the Federal Communications Commission, airspace regulated by the Federal Aviation Administration, and bulk power regulated by the Federal Energy Regulatory Commission. The result was untold billions in forgone wealth. Resilience was not built—brittleness was, such as in cybersecurity vulnerabilities and single points of failure in energy networks. The latter is illustrated by the Colonial Pipeline ransomware saga, which forced the shutdown of a key East Coast fuel supply line, triggering widespread gasoline shortages and dry pumps at stations across the region. Markets are essential for figuring out not just which business models scale, but also how risks are discovered, understood, and contained.

Entanglement and misalignment

OpenAI’s manifesto and other plans (including Biden-era initiatives and the Trump/Blackburn proposals) tend to emphasize centralized coordination and entanglements that libertarians have long resisted, including research subsidies, federally backed testbeds, standards boards, and public-private partnerships. These approaches tend to artificially elevate a handful of dominant firms. The new blueprint unabashedly calls for a “new industrial policy” that uses “government’s existing toolbox for aligning public and private activities: research funding, workforce development, market-shaping tools, and targeted regulation.”

OpenAI’s call for “building a resilient society through accountability, alignment, and management of frontier risks” exhibits little appreciation that industrial policy itself, alongside militarization, can generate the very non-alignment it purports to solve. By overriding market discipline, it crashes through the guardrails that actual capitalism imposes. The historical record shows that very little genuine capitalism prevails in large-scale enterprise once federal involvement takes hold. That can be seen in the recent federal backing of the broadband and the semiconductor industries. The Infrastructure Investment and Jobs Act provided $42 billion for the Broadband Equity, Access, and Deployment (BEAD) program. And last year, the Trump administration decided to take an equity stake in Intel.

Doubling down on failed platitudes and failed progressivism

“The Case for a New Industrial Policy” is more than a contemporary misdiagnosis and monopolistic, statist prescription. It reflects a striking lack of awareness of the progressive policy failures that continue to undermine prosperity today. OpenAI claims that “following the transition to the Industrial Age, the Progressive Era and the New Deal helped modernize the social contract for a world reshaped by electricity, the combustion engine, and mass production.” The report presents these transformations as unambiguous successes, but these same interventions entrenched bureaucracy, distorted markets, and created fiscal and regulatory burdens that persist in a $7 trillion budgetary apparatus. These interventions grease the skids for further government expansion anytime an economic shock occurs.

In an era now witnessing everything from calls for government-owned grocery stores to the effective conscription of doctors by single-payer health care, echoes of Stuart Chase are unmistakable. As he asked in 1932’s A New Deal, “Why should Russians have all the fun of remaking a world?” The technocratic impulse endures, now reframed for the AI age, with the explicit assumption that “we have to steer” it.

This doubling down on progressivism includes a call to “[m]odernize the tax base.” The idea here is not to reduce tax burdens for an AI world that will be wealthier and less in need of government. Rather, it is to shift taxation on labor to corporate profits and capital gains. Ultimately, the purpose of such proposals is to sustain programs like Social Security, Medicaid, SNAP, and housing assistance, all of which already need reform and deconstruction.

A custodial attitude

Invoking a newfound “Right to AI,” the proposal reveals an unnerving caretaker attitude toward the public, emphasizing managed access and redistribution. References to “mass efforts to increase global literacy, or to make sure that electricity and the internet reach remote parts of the globe” are used to justify free or low-cost provision of a baseline level of AI capability.

The tech sector’s longstanding flirtation with a Universal Basic Income gains new license in OpenAI’s proposal for a business- and government-seeded “Public Wealth Fund” that would give every citizen a stake in AI-driven economic growth regardless of market participation.

The plan escalates the wrong-headed fusion of workplace benefits and social policy (already grievously exacerbated by the Affordable Care Act). That is accomplished through proposals for efficiency dividends, whereby AI-driven gains are converted into expanded worker benefits. These include increased retirement contributions, greater employer coverage of health care costs, and subsidies for childcare and eldercare.

OpenAI’s call for portable benefits “not tied to a single employer” would, in another context, represent a move toward greater flexibility and reduced dependence on government. Here, however, portable benefits are framed as part of an expanded system of managed access to health care, retirement savings, and skills training. This suggests not decentralization and privatization, but a broader administrative overlay.

AI, in other words, is being invoked not to enable the dismantling of the welfare state through rising prosperity, but to justify its expansion. Rather than illustrating how AI could reduce reliance on programs like unemployment insurance, SNAP, Social Security, Medicaid, and Medicare, the proposal calls for making them more expansive, more responsive, and more deeply embedded.

In proposing complex metrics to gauge job quality, unemployment, and disruption, one gets the disconcerting sense that OpenAI envisions a world in which workers are increasingly sidelined. The report calls for policymakers to “define a package of temporary, expanded safety nets (e.g., expanded or more flexible unemployment benefits, fast cash assistance, wage insurance, training vouchers)” triggered by newly created metrics of disruption. When those metrics exceed a pre-defined threshold, says OpenAI, “support would scale up; as conditions stabilize, it would phase out.”

How often, in practice, do government social programs actually phase out? It is difficult to take such assurances seriously.

The pedicure and backrub economy

The custodial attitude reflects a diminished vision of American self-reliance and individualism arriving, notably, at the nation’s 250th anniversary of independence. The portrayal of individuals as dependent wards reaches its logical conclusion in OpenAI’s embrace of a “care and connection economy,” as if humanity’s primary characteristic is its sickliness, helplessness, and inability to care for its own offspring.

In a section titled “Pathways into human-centered work,” policymakers are encouraged to expand roles in childcare, eldercare, education, health care, and community services as destinations for workers displaced by AI, supported by government-built training pipelines.

The result is a strangely constrained vision of human potential, and one in which large segments of the population are redirected into service roles. Ironically, many such roles are themselves increasingly subject to automation. The vision edges toward something more troubling: a tacit suggestion that individualism was always overstated and that collectivist management is and always has been the natural endpoint of social organization. Human ambition, however, amounts to more than resignation to a pedicure-and-backrub economy.

Democracy misconstrued

In pursuing an “open economy with broad access, participation, and shared prosperity,” the proposal asserts that “[a]voiding a concentration of wealth and control will require ensuring that people everywhere can use AI in ways that give them real influence at work, in markets, and through democratic processes.”

Yet while invoking democracy, the framework assumes collectivism as a baseline. It conflates the relative contributions of markets and state action to risk and innovation, while relying on the aforementioned coercive mechanisms—grants, partnerships, and incentive programs—that are inherently selective and prone to capture and cartelization.

The plan’s nod to checks and balances is welcome but difficult to take seriously when paired with calls to reimagine the social contract through coordinated, large-scale intervention “mediating between capital and labor, and encouraging broad distribution of the benefits of technological progress.” The result is not democratization but managed societal participation and engagement under administrative state oversight.

The disentanglement imperative

Today’s entanglement blueprints threaten to calcify into permanent bureaucratic and oligarchic influence, even though framed as light touch. The emphasis on federally driven workforce development, safety pledges, equitable distribution, and more risks creating new regulations and sub-regulatory guidance that will be difficult or impossible to unwind.

While startups receive rhetorical support, smaller and more agile innovators will face disproportionate burdens from compliance requirements, reporting obligations, and pseudo-democratic mandatory participation frameworks. Dissident researchers operating outside dominant narratives (e.g., climatologists not on board with climate alarmism) face additional barriers, particularly given their exclusion from subsidized ecosystems and federally linked research environments.

The proposal for a “distributed network of AI-enabled laboratories” across universities, community colleges, hospitals, and regional hubs raises further concerns about indeterminate ownership, distorted property rights regimes, and embedded federal influence. In practice, these new governance layers will operate without clear legislative authorization. Furthermore, expertise is not reliably present, let alone concentrated, within federal agencies, yet their role would expand. With this malign infrastructure, future administrations disinclined toward free enterprise could then easily reshape AI policy in ways that reflect radical political preferences, secure ideological conformity, or advance social engineering goals while maintaining a veneer of neutrality.

Disentanglement, not alignment

Imagine instead a sustainable governance philosophy focused on not steering the private AI sector, one that forbids federal deployments and subsidies, rejects rigid industry-wide standards, and creates a framework capable of resisting the embedding of ideological preferences into technology. With regard to risk mitigation, private insurance receives only passing mention in the OpenAI plan, but its role—along with contractual accountability and other competitive market mechanisms that contain risk—is undermined when policy is pursued, as the blueprint recommends, “collectively, at scale.”

A genuine alignment policy would reject industrial policy outright and constrain Washington first and foremost. The federal government should be prohibited from picking winners and losers. Thus, it should be prohibited from directing private-sector deployment or subsidizing preferred technologies. Federal agencies should reduce their own AI deployments, limit contractor indemnification, and avoid inappropriately expanding the role of federal technical bodies or, worse, building a “global network of AI Institutes.”

The use of AI in warrantless surveillance sweeps also must be prohibited. The industrial policy’s call for systems “designed to support investigation or intervention under clearly defined legal or safety conditions” is especially fraught at a moment when those conditions remain deeply contested as Congress debates reauthorizing the Foreign Intelligence Surveillance Act. Relatedly, the House Judiciary Committee has recently found that some federal agencies have weaponized AI to censor speech. The committee’s report alleges censorship pressures from the White House and federal agencies, including the National Science Foundation, and State Department-linked disinformation initiatives targeting both tech platforms and speech about COVID-19, elections, and political content. We seem already to have forgotten that in 2022 the Department of Homeland Security itself erected a so-called Disinformation Governance Board. Today, federal funding streams involving entities such as USAID and the National Endowment for Democracy have raised concerns about NGOs and contractors shaping elections, governance, and disinformation narratives in ways that can affect speech. Taken together, these dynamics underscore how easily governance frameworks in AI can evolve into instruments of control.

Conclusion: Misconstruing capitalism

Calling for even non-governmental institutions to join the effort, OpenAI argues that industrial policy should align public and private activity through research funding, workforce development, regulation, and market-shaping tools, reinforced by procurement and investment. But these approaches distort incentives, enable rent seeking, and entrench well-positioned interests.

OpenAI acknowledges that capitalism has historically raised living standards and expanded opportunity, yet still insists that industrial policy must intervene when markets fall short. The pattern is familiar: praise markets in theory, override them in practice. In its own words, capitalism remains “an effective system for translating human ingenuity into shared prosperity.” Still, OpenAI regards capitalism as needing correction through coordinated state action when new technologies introduce risks and opportunities.

OpenAI assumes that political systems can capably manage what markets allegedly cannot. In reality, AI systems will differ profoundly depending on whether they emerge from accountable market processes or from a coercive hybrid of state direction and private enterprise.

The claim that “sharing economic gains more widely” should be a goal of AI policy is also misleading. It is precisely capitalism—through competition, innovation, and diffusion—that expands opportunity and raises living standards. Policy prescriptions that reinforce state centrality and enable coercive utopian experimentation would undermine shared economic gains.

AI’s central dilemma is not misalignment by accident or neglect, as the leading blueprints maintain. It is misalignment by design—the deliberate entanglement of government power with private enterprise, the elevation of statism and collectivism “at scale,” over capitalism and competitive disciplines.

Further reading

About the author

Clyde Wayne Crews Jr. is the Fred L. Smith Fellow in Regulatory Studies at the Competitive Enterprise Institute. His work explores the impact of government regulation of free enterprise. He is the author of the annual report, Ten Thousand Commandments: An Annual Snapshot of the Federal Regulatory State.