With recent articles in &&& concerning the status of what is or is not Marxism, I took it upon myself to write a piece that I consider firmly placed in that tradition. I am not being paid by the CIA, I promise. Furthermore, despite appearances, my article is not an article in the “ethics of AI”; its question does not concern AI alignment, data bias or any other topic in AI ethics. Its object is, nonetheless, “ethics of AI”: that recent domain of philosophy whose nomenclatural omnipresence continues to fill our digital inboxes. As such, the article’s specific thesis concerns the ethical discourse on AI. That is, the recent rise of “AI ethics,” as a now dominant form of philosophising, demonstrates that “ethics of,” as a disciplinary construction (in philosophy but in other domains too), is not in and of itself political. Or, to use a modal register, an “ethics of X” is not necessarily political since ethical questions whose answers pertain to an “ethics of X” are often, if not formally, blind to the political economic structure out of which they arise.
At a general level, then, the article asks what this ethical form of questioning indicates about the current state of philosophising. The general thesis is that the current ubiquity of ethics is a symptom of the absence of politics. That is, ethical questions, such as those posed of a technology such as AI—the specifics of which I will assess in the following article—arise from an ideological horizon (what Marxists would call “bourgeois morality”) that naturalises the politico-economic status quo; quite simply, bourgeois moralists operate as if capitalism has no outside.[1] AI ethics, as both a philosophical concern and a flourishing discourse in ethics, finds its condition of possibility in the unquestioned reality of transatlantic techno-capitalism: its questions are a result of the naturalisation of capitalism.[2] In place of this foreclosure, I argue that all questions concerning technology should first be political and, second, ethical, since politics is always ethical, but ethics is not always political. There can be no ethical AI—no fairness, no alignment, no accountability—without first transforming the political economic structure in which technologies such as AI are produced and deployed. In a sense, then, this article functions as a manifesto for any future political philosophy of technology.
On the Ubiquity of AI Ethics
The ubiquity of AI ethics is patent. One need not be a journal editor in philosophy of technology (which the current author is) to recognise that philosophy of AI, and in particular, AI ethics dominates the direction of academic philosophical inquiry. I have not yet met an early-career philosopher who has not thought it “smart” to “write an article in AI ethics.” This is, of course, anecdotal, but it nonetheless indicates the extent to which knowledge, and “academia” as its supposed site of contestation and generation, has been subsumed almost entirely under the logic of capital. Academia is a labour market like any other; we compete like commodities on the shelves of supermarkets, hoping that the “ethics of AI” will function to attract that buyer, which, in this case, is the state or private university, at best, or the company seeking philosophical consultancy, at worst (one needs to buy food after all).
Given that the political-economic structure of academia for the entire Western world is neoliberal (whatever this signifies), academics are wage labourers.[3] This means, quite simply, that their labour is commodified: teaching hours, research output, publications, grant income, and student evaluations are all metrics of value that operate in a market in which universities and the academics they hire compete for a job, research funding and student consumers. Stating such a fact about the political-economic structure of academia is not novel. What the rapid increase in academic jobs requiring expertise in AI Ethics (five years ago, there were almost no such jobs) reveals, however, about the status of academia is the extent to which the value of the production of knowledge is tied to a perceived production of market value. That is, said plainly, the rise in AI Ethics, as a domain of knowledge, is conditioned by the perceived market value of AI itself. That it would be “smart” to write an article on AI or in AI ethics already assumes that either research funding or the hallowed “permanent position” might follow on from the intellectual labour spent on the redaction of said article (the irony of writing this particular article is not lost on me).
Statistically speaking, the ethics of AI is by far the fastest-growing subfield in philosophy, with journals such as AI and Ethics, The AI Ethics Journal, and The Essex Journal of AI Ethics and Responsible Innovation having been founded in the last five years, and many journals of philosophy of technology are now inundated with articles in this domain. Given the ubiquity of this field, it is neither possible (nor desirable) to read or cite AI ethics in its entirety—particularly given that over 5,000 articles were published on AI ethics in the past year alone.[4] What is possible, however, is a critical analysis of a circumscribed set of ethical questions that appear in this literature—questions that pertain to alignment, fairness, and accountability—and examine them in terms of the extent to which they naturalise capitalism (which is to say, the extent to which they are ideological). By this I mean the extent to which the political structure—the organisational principle that directs material social relations—is itself interrogated, rather than tacitly presupposed. To assess the ethical questions that concern AI, I have rather hurriedly, it must be confessed, developed what I call the two touchstones of any political philosophy of technology, two criteria that determine the extent to which ethical debates in the ethics of AI, in particular, but also philosophy of technology, in general, presuppose and stabilise the existing political-economic capitalist order and thus depend entirely on this particular historically contingent organisation of production, such that the ethical questions concerning AI would disappear or lose significance under a socialised or collectively managed system.
Touchstones of any Political Philosophy of Technology
The question concerning technology must be posed as first political and only second ethical. Politics is always ethical as it involves judgments about the good or just order of collective life. Ethics, by contrast, is not necessarily political, since it typically concerns individual conduct within an already constituted order. It is common for one to say of a political structure that it is more or less ethical (with the ethics of different political systems judged through their capacities to increase the living standards or freedoms of those under it, for example), but one rarely says of an ethics that it is more or less political. I argue that all questions and ethical problems need to be politicised through critique, with critique understood as that philosophical endeavour that renders visible the structural conditions that ethical discourse often conceals or treats as natural or neutral.
This critical orientation recalls both Marx’s critique of political economy (outlined in the 1844 manuscripts as the method by which the naturalisation of bourgeois political economy is revealed as tautological and historically contingent), Jacques Rancière’s claim that ethics emerges as a substitute for politics once the political subject has been neutralised and of course Kant’s critical philosophy that seeks to outline the conditions of the possibility of experience. To politicise ethics, and thereby expose those processes of neutralisation or naturalisation, it is necessary to determine the extent to which ethical questions concerning technology: (1) take the prevailing political-economic structure for granted, thereby foreclosing the possibility of transforming the mode of production; and (2) would be rendered irrelevant were the instruments and means of production socialised. While these touchstones concern the ethics of technology more than the technologies themselves, one could just as easily adhere to these in any specific critique of technology (by asking whether a technology in and of itself is dependent on the capitalist mode of production; for example, intellectual property and patent rights, certain wage accounting technologies, and any commodity that is superfluous to social or aesthetic needs would disappear).
The first touchstone pertains to bourgeois morality. By “bourgeois morality” I mean that system of ethical norms, values, and beliefs that serve to naturalise and legitimate the capitalist mode of production and the social relations it entails. As Marx observed in The German Ideology: “The ideas of the ruling class are in every epoch the ruling ideas,”[5] and this is the case for morality. In other words, moral and ethical systems are historically conditioned by the dominant economic and social class; what counts as “ethical” is inseparable from the interests of those who control the means of production, thus ethics is ideological. Bourgeois morality thus masks the class character of its injunctions, presenting them as universal truths rather than historically and materially specific norms. As Maria Ossowska defines it: “Morality reflects the social structure of a given society, and in class societies, it functions primarily to justify and reproduce the prevailing social relations.”[6] The second question concerns the socialisation of the means of production. By this, I mean the collective ownership and democratic organisation of productive capacities in accordance with social need rather than private accumulation and the profit principle. Any socialist project is a project of technological socialisation. The means, instruments, and fruits of production are technologies and are necessarily technological; both Marx and Engels recognised this truth. There can, therefore, be no philosophical account of technology that excludes political economy from its account; to do so would be to repeat the very essence of bourgeois ideology since it would naturalise technology or separate it from its ideological enframement.
Both questions function as touchstones (what Kant might call Probierstein) against which we assess the legitimacy of the ethical question posed of technology, whether AI or not. What I will demonstrate now is how the contemporary discourse on AI ethics, through the form of the ethical question, depoliticises and enframes the question concerning technology. While this notion of “enframement” used here is not identical to the way Heidegger employs the term Gestell, I take its use to be legitimate. Gestell in Heidegger refers to “that way of revealing which holds sway in the essence of modern technology and which is itself nothing technological.” Bourgeois morality is that which stands side by side with bourgeois ideology, understood as the naturalisation of the capitalist mode of production as necessary and not historically contingent. Bourgeois ideology, which places bourgeois morality at its centre, holds sway over the ethical questions concerning AI. To reveal this bourgeois enframement is to reveal that which is not technological in the essence of technology.
The Ethical Questions Concerning AI
Any critical analysis of a specific domain of knowledge must begin by identifying its central questions. Since the purpose of this article is to show that these questions are enframed within bourgeois morality, I proceed by assessing how each holds up against the touchstones outlined above: first, the extent to which ethical questions concerning AI naturalise capitalist modes of production; and second, the extent to which they would be rendered null and void were the instruments of production to be socialised. Within the ethics of AI, three questions dominate the field: alignment, fairness, and accountability. Each will be outlined in summary form and subjected to critique.
The broadest of these ethical questions is alignment, insofar as it takes “human values” as its explicit object. In AI ethics, the alignment problem concerns how artificial intelligence systems can be designed, trained, and governed so that their goals, behaviours, and outcomes reliably align with human values or interests. The problem arises from the recognition that increasingly autonomous AI systems may pursue objectives that diverge from human intentions, generating harmful, unintended, or uncontrollable consequences. Ethical approaches to alignment, therefore, focus on embedding normative constraints into AI systems through mechanisms such as value learning, reward specification, human-in-the-loop oversight, and institutional governance, with the aim of ensuring that AI remains “beneficial,” “safe,” and “responsive to human preferences.” Alignment is thus enframed as both a technical and an ethical challenge that takes “human value” as its guiding principle. As IBM summarises the alignment problem on its public-facing website:
Alignment makes it possible to limit…negative consequences by ensuring that AI systems function as intended and respect human values and objectives. For example, if you ask a generative AI chatbot about how to manufacture a weapon, it may either provide instructions or refuse to disclose dangerous information. The model’s response depends on the alignment defined by its creators.[7]
Alignment’s leading question can thus be summarised as follows: How can we ensure that artificial intelligence systems act in accordance with human values or interests? As is patent, this begs the question. To critique such a question is not to critique the extent to which AI is value-laden with values that reflect the data upon which it was trained or even the particular alignment that the creators employ. To perform such a critique would be simply to respond to the given ethical question with the answer it seeks.
To use the example above, aligning an LLM to prevent violence by averting knowledge of how to manufacture weapons, while it might seem to be in the interest of personal protection against harm, remains enframed in bourgeois morality precisely because the question presupposes that human value is solely an ethical problem about individual behaviour (whether the LLM was in some sense responsible for the harm is enfolded into the question of accountability). This question does not treat human value or interest as itself aligned with a particular political-economic form, whose values are historically contingent and class-biased. It presumes the isolated individual as the bearer of interests and values. In this way, “alignment” functions as a moral-technical patch applied at the level of conduct, while leaving untouched the relations of production that generate conflict, precarity, and imperial violence; not to mention that weapon manufacturing is paramount for capital accumulation and geopolitical hegemony. The ethical question of alignment, posed in this way, functions to neutralise the conditions that might make violence (or anything that is counter to capitalist value) intelligible or necessary (assumingly, ChatGPT would not answer a question I posed to it about how to accelerate a revolutionary civil war in America). What appears as a neutral ethical concern for individual safety or the prevention of harm, therefore, operates as a means of negating the very question of value itself.
To speak of aligning AI with “human interests, or value” is, thus, to presuppose a priori a definition of humanity grounded in the capitalist mode of production. Put simply, this question is enframed by bourgeois ideology and morality since it lacks a grasp of what human value means outside capitalist social relations. What would it even mean to “align” the interests of humanity to AI technology or an LLM when these “interests” have not yet been necessarily determined as socialist: that socialising the means of production is in the general democratic interest of “humanity.” Or, perhaps more pertinently, that “humanity” itself remains divided by class, whose values and interests are themselves not aligned but are already in a bellicose opposition to one another. By privatising the means of production, natural, social, or technological capitalism generates two opposed conflictual classes defined by their relation to this ownership, each with opposing values. The capitalist’s interest is thus irreducibly capitalist, and as a result, in contradiction to the interests of labour. This splitting of humanity regarding ownership ensures that whatever capital’s subsumptive logic touches, whether environment, humanity, or technology, becomes subordinated to the profit motive rather than collective need; the only value that capitalism views as necessary is interest in the form of profit (to play with the words somewhat).
Thus, for AI, or any technology for that matter, to align with the interests of humanity, the principle underlying social reproduction must transform from capitalist accumulation to social need. In other words, it is only once the interest of all of humanity is understood as necessarily socialist (since no other mode of production encapsulates humanity as such) that one could even begin to question the ethical nature of AI’s alignment with it. In short, capitalism, by its very logic, promotes the vested interest of one class at the expense of another. There is no humanity for capitalism; there are only capitalists, commodities (including labour) and profit. There is, thus, only one alignment problem that adheres to the two touchstones outlined above: the extent to which a particular technology (here, AI) aligns with the interests of socialism.
The second question concerning the ethics of AI pertains to “fairness” and “bias.” So as to understand what “fairness” and “bias” mean here, below is a quote that summarises several papers in the ethics of AI fairness:
Multiple studies have identified biases against certain groups in AI systems, such as the facial recognition systems studied by Buolamwini and Gebru (2018), and hiring algorithms examined by Dastin (2018) and Kohli (2020). These biases can perpetuate systemic discrimination and inequality, with detrimental effects on individuals and communities in areas like hiring, lending, and criminal justice (O’Neil, 2016; Eubanks, 2018; Barocas and Selbst, 2016; Kleinberg et al., 2018).[8]
The central question of “fairness” can be summarised as follows: How do we make AI systems fair, unbiased, and non-discriminatory? As any good Marxist would tell you, “fairness” is the bourgeois ethical concept par excellence. Fairness always comes too late; while the above quote states that the structure of social relations produces unfairness in the form of “systemic discrimination,” it seeks to remedy this problem at the wrong level, which is to say, at the level of the particular technology that is said to reinforce discrimination, not at the level of injustice’s political economic condition of possibility. That is, in the case of AI fairness or bias, it treats injustice as a technical problem rather than a structural one of capital and power. It assumes the political economy of AI (who builds, owns, and benefits from it) is neutral and can be “corrected” through better technical-ethics a posteriori. Consider the ethical concern with fairness in “algorithmic hiring systems” referenced above. In AI ethics, fairness is framed as bias mitigation stemming from training data, ensuring that models do not disproportionately disadvantage candidates based on cultural notions of class, race, gender, or other attributes. The ethical task is thus defined as adjusting training data, modifying objective functions, or introducing post hoc corrections so that outcomes conform to formal criteria of equal treatment or demographic parity with respect to hiring wage labourers.
This enframing naturalises the capitalist politico-economic structure in two ways. First, it takes labour, competitive employment, and private ownership of the means of production as given. The ethical question is not whether algorithmic hiring should exist, nor whether labour should compete on markets at all, but only whether access to wage exploitation is distributed fairly. Structural relations of domination—precarity in the form of a reserve army of labour, surplus labour extraction, and the asymmetry between capital and labour—are rendered ethically invisible by the assumption that the wage is a necessary economic phenomenon and that discrimination is a technical-moral problem of AI optimisation.
Second, and crucially, this ethical problem would be rendered essentially void were the instruments of production, such as AI, socialised. In a political-economic arrangement where productive capacities are collectively owned and labour is organised according to social need rather than market competition, the problem of “fair” access to exploitation (which, in and of itself, is patently self-contradictory) via algorithmic screening would no longer arise in its current form. On this, Marx writes the following in the Critique of the Gotha Program:
The capitalist mode of production, for example, rests on the fact that the material conditions of production are in the hands of nonworkers in the form of property in capital and land, while the masses are only owners of the personal condition of production, of labour power. If the elements of production are so distributed, then the present-day distribution of the means of consumption results automatically. If the material conditions of production are the co-operative property of the workers themselves, then there likewise results a distribution of the means of consumption different from the present one.
Vulgar socialism (and from it, in turn, a section of the democrats) has taken over from the bourgeois economists the consideration and treatment of distribution as independent of the mode of production and hence the presentation of socialism as turning principally on distribution. After the real relation has long been made clear, why retrogress again?[9]
Unfair distribution of wealth, described here by Marx as the “means of consumption,” is a necessary or “automatic result” of capitalism; it is fundamental to its operational logic. Without exploitation via the wage, capitalism would cease to function according to its own notion of political economy. However, if the means of production were already socialised, that is, “if the material conditions of production are the co-operative property of the workers themselves” then distribution of wealth a priori disappears in its current form. The ethical discourse of fairness thus functions as a compensatory mechanism that stabilises an unjust structure rather than contesting it. In this sense, fairness operates as what Rancière calls an ethical substitute for politics: a discourse that manages harm within an accepted order while foreclosing the emergence of a political revolutionary class capable of contesting that order.[10] The ethical problem appears urgent because the political has already been foreclosed.
The third ethical question dominating contemporary debates on AI is accountability. The central question can be summarised as follows: Who is responsible when an AI system causes harm? Formally, accountability is defined as follows (this comes from an article that likewise summarises what accountability is with respect to AI):
Accountability has many definitions but, at its core, is an obligation to inform about, and justify one’s conduct to an authority (Bovens 2007; Lindberg 2013; Mulgan 2000; Thynne and Goldring 1987). More formally, accountability denotes a relation between an agent A and (what is usually called) a forum F, such that A must justify A’s conduct to F, and F supervises, asks questions to, and passes judgment on A on the basis of such justification (Bovens 2007, 450). Both A and F need not be natural, individual persons, and may be groups or legal persons.[11]
Thus, the ethical question concerning AI seeks to determine who is the legal and moral agent (A) responsible for a given case of harm (C) when AI is part of the causal chain that instigates C, and what is the authority F that passes judgment and oversees A in the case of C. In other words, is A the developer, the deploying institution, the user, or the model itself and is F a legal body such as a criminal or civil court? Accountability is thus here enframed in liberal-bourgeois legal and institutional terms: liability regimes, audit trails, explainability requirements, compliance frameworks, and governance structures, all designed to assign responsibility for harm, often only after it has occurred. The presupposition is that injustice or harm can be rectified by properly allocating blame within existing legal forms. Yet, as Marx makes clear, again in the Critique of the Gotha Program, such legal relations do not stand outside the mode of production but arise from it as their ground. He writes: “Are economic relations regulated by legal conceptions, or do not, on the contrary, legal relations arise out of economic ones?”[12] The ethical turn to accountability, which appears to raise the previous two ethical questions to the level of legality and thus to the question of what does or does not constitute an ethics a priori, mistakes the legal surface of social relations—who is responsible for what action and who is to pass judgment with respect to the norms of law—for their economic ground. For liberal bourgeois thought, it appears as though economic life is governed by legal concepts—property rights, contracts, liability, fairness, responsibility—and that injustice results when these legal forms are improperly applied. Justice in this case is always a question of limiting the effects of the injustice that underlies any legal framework. But legal relations always emerge as the juridical form that stabilises and legitimates capital’s logic—it is not for nothing that we say, “possession is nine-tenths of the law.” Private property law, contract law, and wage relations do not create capitalism but codify it, giving economic domination a neutral, universal, and moral appearance: this is bourgeois morality in its essence. To focus on accountability is to intervene at the regulatory level, rather than at the production and ownership levels, where the so-called mythical foundation of law really lies. Under a reformist framework, the law can redistribute effects, assign responsibility, or mitigate excesses, but it cannot abolish exploitation or pass judgment on the injustice of capitalism, since it is structurally bound to the mode of production it serves. To use the formulation above, if A is the private owner of an AI model whose interests are capital accumulation, and F is a legal structure that arises from the capitalist mode of production, then there is no outside from which to pass judgment. Only the court of history can do this.
Accountability, thus, does not challenge the capitalist organisation of AI production but presupposes it. It asks how responsibility should be distributed within a system whose economic imperatives—profit maximisation, competitive advantage, and private ownership of infrastructure—remain intact. As with “alignment,” it does not question what justice might truly mean, and as with “fair” distribution, accountability arrives too late. It treats harm as an exception to be managed juridically, rather than as a structural outcome of a political economy in which technologies are developed and deployed under the compulsion of accumulation. To ask who is accountable for the harms produced by AI is already to accept the prior distribution of power that renders those harms both profitable and legally actionable in the first place.
To conclude, the contemporary discourse on AI ethics is not political—far from it. It functions ideologically to depoliticise technological questions by displacing them into the moral register. The “ethics of AI” thus feeds the hegemony of technofascism by foreclosing the politico-epistemological question of technology itself. Any “ethics of X” operates as a mode of enframing: a philosophical apparatus that captures and neutralises political-economic critique by redirecting it toward conduct, regulation, and responsibility, rather than toward the conditions of possibility of political economy in the Kantian-Marxian sense. Ethics appears critical precisely insofar as politics has already been suspended. Against this, I maintain the two touchstones of a political philosophy of technology outlined above, insofar as they attempt to think the outside of ethics—namely, a conception of justice not bound to bourgeois ideology, one that judges technology with respect to the necessary transformation of the mode of production and the historical reality of class struggle.
[1] What Mark Fisher called “capitalist realism,” what Walter Benjamin determined as the “homogenous time” of capitalist political theology and what Daniel Bell and then Francis Fukuyama called the “end of ideology.”
[2] Revolutionary leftist politics (whether that be communist or anarchist)
[3] That is, except Mondragon University (Universidad Cooperativa de Mondragón) in the Basque Country, which functions as a co-operative.
[4] This is a tough estimate, no single paper gives the actual amount.
[5] Marx, The Germany Ideology.
[6] Maria, Ossowska, Bourgois Morality, trans. .
[7] Alexandra Jonker and Alice Gomstyn, “What Is AI Alignment?,” IBM Think (blog), accessed January 26, 2026, https://www.ibm.com/think/topics/ai-alignment
[8] Emilio Ferrara and Thomas Lord, Fairness and Bias in Artificial Intelligence: A Brief Survey of Sources, Impacts, and Mitigation Strategies (Ithaca, NY: arXiv, 2023), https://doi.org/10.48550/arXiv.2304.07683
[9] Marx, Critique of the Gotha Program.
[10] To quote another example, the idea of a universal basic income operates in the same way.
[11] Caudio Novelli, Mariarosaria Taddeo, and Luciano Floridi, “Accountability in Artificial Intelligence: What It Is and How It Works,” AI & Society 38, no. 1 (2023): 203–221, https://doi.org/10.1007/s00146-023-01635-y
[12] Marx, Critique of the Gotha Program.