US lawmakers split on AI regulation amid safety fears
The potential for artificial intelligence to become dangerously uncontrolled affects national security, the economy, and the public, requiring government intervention.
The plan - and there is not yet a plan, only the demand for one, which is itself instructive - requires that the practice of governing something by working alongside it be replaced by the practice of governing something by declaration issued in advance of the knowing. Congress returned to Washington this week, after its recess, to warnings that artificial intelligence might slip its controls, and found itself, as of Monday, without agreement on what an intervention would even consist of. This is not a failure of will, whatever the newspapers say. It is a symptom of the kind of thing artificial intelligence is, meeting the kind of thing legislation is, and the two not fitting.
The crisis arrived on a timeline that assumed institutions capable of processing it at a speed they last achieved in the age of the telegraph, when a message could cross the continent faster than a committee could be convened to discuss what to do about it. The institutions, naturally, processed it at the speed they are equipped for, which is the speed of a previous era - in this case, the speed of a Monday. Congress returned to Washington from recess, as it does, on the calendar of a nineteenth-century agrarian legislature, to confront a force that had not paused for recess at all, that had in the intervening weeks trained on more material, compressed more computation, and extended its reach into more decisions than the sum of what the human institutions of governance had managed in the same span. The mismatch is not a metaphor. It is the story.
On the floor of a fulfillment center outside Columbus, Ohio, a woman named Loretta picks orders for eleven hours a shift, and the pace she walks is not set by her supervisor. It is set by a piece of software that scores every second she is not moving toward a shelf. That system has already spun out of anyone’s control but the company that owns it. The debate happening in Washington this week will decide whether anyone in government even has the authority to ask that company a question about it.
The energy in artificial intelligence moves from researcher to laboratory to product through a circuit of capital risk and iterative failure: a firm spends money it may lose, tests a model against a market that may reject it, and adjusts on the feedback of profit or loss. Congress, returning to Washington this Monday from recess, proposes to intervene somewhere in that circuit. It has not yet decided where, and this is the detail worth pausing on, because the location of the contact point matters more than the fact of contact itself.
Every participant in this debate accepts that the question before the Congress that returned to Washington on Monday is one of calibration - how much intervention, what kind, how soon - against a technology that is described, again and again, as something that might “spin out of control.” None asks when artificial intelligence became a weather system rather than a commodity, or who benefits from a debate that treats the direction of a specific set of capital investments as an autonomous natural force requiring only management. The grammar of the warning is the story.
Paterson-style
The energy moves from researcher to user through the circuit of venture capital, algorithmic iteration, and market adoption. The proposed intervention breaks the circuit at the point of regulatory oversight, inserting a bureaucratic node that substitutes political judgment for market feedback.
Your argument rests on the premise that artificial intelligence is a weather system - an autonomous force beyond steering - that requires management by experts. You identify the danger not in the machinery itself, but in its capacity to spin out of control, relocating the threat from the boardroom to the machine. This is a compelling diagnostic of the political impulse. It is true that the language of “control” serves ideological work by depoliticizing corporate strategy. When executives frame their pursuit of market share and rent extraction as an inevitable technological tide, they invite the state to become the dam. You are correct that the current debate treats capital investment as a natural phenomenon rather than a series of human decisions. This is the strongest point of your position: the recognition that the crisis of AI is, in part, a crisis of attribution, where corporate power hides behind the veil of technical inevitability.
I concede that the framing of AI as an uncontrollable force is a rhetorical device used by those who benefit from state protectionism. However, your framework assumes that the solution to this misattribution is better regulation by “experts” and “regulators.” This is where the circuits diverge. You see the blockage as a failure of governance; I see it as a failure of transmission. The error is not that the machine is too powerful, but that the feedback loops which normally constrain power have been severed by the very intervention you propose.
Consider the circuit of innovation. Energy enters as capital and intellectual labor. It flows through the competitive pressure to produce useful, efficient, and affordable services. The consumer provides the feedback: adoption or rejection. This is the Long Circuit. When a technology fails, the circuit breaks, and the capital is withdrawn. When it succeeds, the capital multiplies. This is not a weather system; it is a biological system of adaptation.
The intervention you endorse - state regulation of AI development - inserts a new node into this circuit. This node does not measure productivity. It measures compliance. The energy that would have flowed through rapid iteration and competitive pressure is rerouted through legal departments, lobbying firms, and regulatory submissions. The result is not a safer machine; it is a slower circuit. The downstream effect is not the taming of the AI, but the consolidation of the platforms that can afford to navigate the regulatory maze.
You argue that Microsoft, Google, and Amazon are extracting rent. I agree. But the extraction is not caused by the AI itself. It is caused by the barrier to entry created by the expectation of state intervention. When the state signals that it will regulate the “weather,” it signals that it will protect the homeowners who have already built their roofs. The small innovator, the garage researcher, cannot afford the regulatory overhead. The large platform can. Therefore, the regulation does not stop the rent extraction; it entrenches the extractor. The circuit becomes blocked not by the technology, but by the cost of compliance.
This is not a novel observation. It is the same pattern that appeared in the early days of the telephone. The energy for the telephone network did not come from the Bell Company’s benevolence; it came from the specific decisions of investors who stood to profit from a connected world. When the state intervened to set rates and dictate infrastructure, it did not make the system more democratic. It made the system more rigid. The “expert” regulators, removed from the circuit of profit and loss, could not distinguish between a necessary innovation and a destructive one. They blocked the path, and the energy stagnated.
The danger of AI is not that it will spin out of control like a weather system. The danger is that it will be controlled by people who do not understand the circuit. When you place the steering wheel in the hands of bureaucrats, you do not stop the car from crashing; you simply ensure that the crash is delayed until the vehicle is too large and complex to maneuver. The feedback mechanism is removed. The regulator does not suffer when the AI fails; the user does. The investor does not suffer when the AI fails; the consumer does.
You seek to manage the weather. I seek to clear the sky. The energy that built the digital age is the same energy that built the industrial age: the specific decisions of specific people, held accountable by the specific consequences of their actions. Any intervention that severs this link does not protect the public; it protects the powerful from the accountability that comes with freedom. The circuit is not blocked by the machine. It is blocked by the man who claims he can control it.
Antonio Gramsci
Every participant in this debate accepts that the regulation of artificial intelligence is primarily a question of technical risk management and market efficiency. None has asked when this became obvious, or who profits from the consensus. The assumption is the story. We are told that AI moves through a circuit of capital risk, that Congress must find the right contact point to intervene, and that the danger lies in the system’s potential to “spin out of control.” This framing treats the technological apparatus as a neutral machine awaiting a competent hand, rather than recognizing it as the material expression of a specific historical hegemony - one that has naturalized the acceleration of capital accumulation under the guise of innovation. The libertarian opponent argues that intervention should be targeted at specific points of failure, such as training or deployment, because the harm is caused by specific actors using the system for specific purposes. This is a sophisticated defense of the status quo, but it rests on a fundamental error: it assumes the circuit itself is open to correction by feedback mechanisms, ignoring that the feedback loop is structured to prioritize the preservation of the dominant class’s intellectual and cultural monopoly.
The libertarian position is strong in its insistence that abstract fears of “spin out of control” are unexamined. It is correct that we must ask toward what end a system spins, through what mechanism, and at whose hands. A model does not become dangerous in the vacuum; it becomes dangerous when it serves the interests of those who control the data and the compute. However, the opponent’s framework diverges from reality in its belief that the “circuit of capital risk” is a self-correcting engine. In our historical moment, the circuit is not correcting; it is consolidating. The “iterative failure” mentioned is not a neutral process of discovery but a method of externalizing risk onto the public while privatizing the gains. When we treat the location of government intervention as merely a technical detail - training versus deployment - we miss the political substance of the struggle. The question is not where to place the hand on the switch, but who owns the machine that pulls it. The hegemonic assumption here is that the market, even a regulated one, is the appropriate arbiter of social value. This is a product of the post-war intellectual project that sought to displace class struggle with managerial efficiency.
Consider the role of the “organic intellectual” in this new technological formation. The current discourse on AI is dominated by a cadre of technical elites who produce knowledge in the interest of capital expansion. They articulate the experience of the corporation as the universal experience of humanity. Their “common sense” is that AI is a tool for optimization, a neutral force that benefits all if only the technical details are managed correctly. This is the intellectual leadership of the new bourgeoisie. The libertarian’s call for targeted intervention at specific points of harm assumes that the “specific actor” causing harm is an aberration, a rogue element in an otherwise sound system. This is the error of the liberal reformer. The harm is not an aberration; it is a feature. The extraction of data, the displacement of labor, and the concentration of cognitive power are the natural outputs of this mode of production. To regulate only the “transmission speed” or the “feedback” is to preserve the engine while tweaking the exhaust pipe.
There is a crack in this hegemonic framework, visible in the growing resistance of the subaltern classes who are subjected to algorithmic management without consent. They do not experience AI as a tool for optimization; they experience it as a mechanism of control. The contradiction lies in the fact that the same system which promises liberation through automation delivers subjugation through surveillance. The counter-hegemonic possibility emerges when this experience is articulated not as a technical glitch but as a political injustice. We must move beyond the debate over where Congress should intervene and ask who is doing the intellectual work for which group. If the organic intellectuals of the tech sector continue to frame regulation as a barrier to innovation, they are serving the interests of capital. But if we can cultivate organic intellectuals from within the working class and the marginalized communities most affected by algorithmic bias, we can begin to construct a new common sense - one that sees AI not as a neutral force of nature, but as a site of class struggle.
The historical parallel of the Fordist assembly line is instructive. Just as the assembly line was not merely a technological advancement but a social reorganization that intensified labor extraction, AI is a social reorganization that intensifies cognitive extraction. The regulation of the assembly line was not just about safety valves; it was about the rights of workers to control the pace and conditions of their labor. Similarly, the regulation of AI must be about the democratic control of the algorithmic apparatus. The libertarian’s focus on “specific actors” and “specific purposes” is a distraction from the structural reality. The apparatus itself is political. To treat it otherwise is to accept the hegemony of the ruling class’s intellectual production. We must not ask where to place the hand on the switch; we must ask who holds the lever, and why they have been allowed to do so without challenge. The crisis in AI regulation is not a technical crisis; it is a crisis of legitimacy. The old framework of market-based innovation is dying, and the new framework of democratic control cannot yet be born. In this interregnum, we see the morbid symptoms of a society struggling to articulate a new common sense.
The Verdict
Where They Agree
The most significant point of agreement is a shared suspicion of the dominant narrative that artificial intelligence is an external, runaway force. Both the libertarian and socialist analyses treat the phrase “spin out of control” as an ideological construct that conceals the decisions of specific actors. The libertarian argues this framing invites clumsy, visible regulation that disrupts efficient market feedback loops, while the socialist argues it relocates blame from the capitalist boardroom to a mystified machine, protecting existing power structures. Their agreement reveals that the debate over AI regulation is, at its core, a debate about the legitimacy of attributing consequences - whether the primary cause of harm is flawed market signals or an unjust mode of production, both sides concur that the “uncontrollable AI” story is a diversion.
both positions rest on a shared, unstated belief that Congressional gridlock is a predictable outcome, not an accident. The libertarian sees it as a failure to identify the correct “contact point” in the circuit, while the socialist interprets it as a functional “paralysis is a policy” mechanism that preserves the status quo. Neither entertains the possibility that Congress might competently design and implement effective, nuanced regulation. This shared skepticism of the state’s capacity for benevolent intervention is a foundational premise that drives their analyses down separate but parallel tracks, each expecting governmental action to be either technically misguided or structurally complicit.
Where They Fundamentally Disagree
The fundamental nature of the market’s “circuit” is the primary fracture. The empirical question is whether the feedback loop of capital risk and consumer adoption is a self-correcting biological system of adaptation or a mechanism that inevitably consolidates power and externalizes harm. The libertarian’s steelmanned position is that the circuit, left unobstructed, produces rapid iteration and error-correction that benefits the public through competition. The socialist’s steelmanned counter is that the circuit is not neutral but is structurally designed to prioritize capital accumulation, making consolidation and worker subjugation not a bug but a feature. Normatively, this hinges on a deeper disagreement about the very purpose of an economic system: is its goal efficient allocation of resources or the equitable distribution of power?
A second irreducible disagreement concerns the relationship between regulation and corporate power. The empirical dispute is whether state intervention entrenches incumbents by creating barriers to entry or can be a tool to dismantle monopolistic control. The libertarian argues with high confidence that regulation inherently favours large platforms that can absorb compliance costs, thus stifling the innovators who would disrupt them. The socialist contends that the absence of robust, democratic regulation is what allows capital to concentrate unchecked, and that the goal of intervention should be to challenge ownership models, not just manage risk. The normative divide is over the state’s role: a necessary evil that should be minimized or a potential vehicle for democratic will that must be seized and transformed.
Hidden Assumptions
- Paterson-style: Assumes that the cost of regulatory compliance inherently and always outweighs its public benefit, a claim that could be tested by comparing innovation rates and market concentration in heavily-regulated versus lightly-regulated tech sectors. If false, and smart regulation could demonstrably increase competition and safety without stifling innovation, the entire argument for minimal intervention collapses.
- Paterson-style: Assumes that market signals (profit/loss) provide faster and more accurate feedback than regulatory oversight, a claim that depends on the specific type of harm. For harms with long-tail consequences or diffuse victims, like algorithmic bias or existential risk, this assumption is highly contestable.
- Antonio Gramsci: Assumes that the creation of a “counter-hegemonic” common sense led by organic intellectuals from the working class is a feasible and effective path to change, a claim that rests on a specific theory of political change. If such a movement fails to materialize or is ineffective, the prescription offers no alternative path to democratic control.
- Antonio Gramsci: Assumes that the primary harm of AI is its function as a tool for class domination, a claim that could be tested against evidence of AI applications that materially benefit disadvantaged groups without concentrating power. The discovery of significant counterexamples would challenge the monolithic view of AI as an inherently oppressive force.
Confidence vs Evidence
- Antonio Gramsci: The claim that “paralysis is a policy” - tagged with high confidence as a deliberate feature of the system - is asserted as a structural inevitability without empirical evidence of Congressional intent or a causal mechanism beyond class interest. This is an overconfidence that presents a interpretive framework as an uncontestable fact.
- Paterson-style: The analogy that AI regulation will inevitably follow the pattern of early telephone regulation - tagged with high confidence - relies on a historical parallel that is suggestive but not definitive evidence for a novel technological context. The confidence level implies a deterministic historical law that may not apply.
What This Means For You
When evaluating coverage of AI regulation, be immediately suspicious of any analysis that takes the “uncontrollable force” narrative at face value without questioning who benefits from that story. The most important question to ask is not if the government should act, but what theory of harm and what theory of change underlie any proposed action. Does the proposal target specific human decisions or attempt to manage an abstract “risk”? Look for concrete evidence about who bears the costs of innovation and who reaps the rewards, rather than accepting broad claims about “progress” or “safety.” Your view on this debate should shift significantly upon seeing data that tracks the number of new, small-scale AI entrants to the market following a major regulatory announcement.