Theses for an ecosocialist critique of artificial intelligence

First published in French at Alencontre. Translated by Adam Novak for Europe Solidaire Sans Frontières.

This text deals essentially with generative AI. The formulation in theses (of unequal length) is not intended to establish certainties, but to facilitate debate through the conciseness of the exposition.

Intelligences and human intelligences

1. What we call intelligence is what enables us to grasp difference, to apprehend the new, to anticipate the possible in the course of events that punctuate time.

2. Intelligence is an emergent product of the non-linear evolution of living things.

Nature makes leaps. Inert things are not intelligent. Symbiotic organisations of plants and fungi communicate and adapt to events without anticipation or consciousness. Intelligence as defined here appears in the animal kingdom, where it takes diverse forms and degrees. In unicellular organisms and organisms without a brain, it merges with the “survival instinct” (survival mechanisms).

3. Human intelligence combines a great capacity for abstraction from a small number of data, sophisticated communication, thought, and a developed spiritual life expressed in complex symbolic achievements, both individual and collective.

4. Homo sapiens identifies regularities and symmetries in its surroundings from earliest infancy, and therefore also what is rare or unusual. Absent in other primates, this aptitude underpins our species’ capacity to classify objects through reason and to penetrate their mechanisms through science.

5. Without human society, without bodies communicating and collaborating, there is neither reflexive intelligence, nor spiritual life, nor consciousness.

The characteristics of our intelligence result both from physical traits (the volume and structure of the brain, bipedalism, the specialisation of the hand, the vocal apparatus) and from the fact that Homo sapiens is a social mammal. The young of our species can survive only thanks to prolonged parental care; we exchange through a complex syntactic language; and our social relationship with the rest of nature is mediated by labour, carried out with the aid of tools. These traits confer upon Homo sapiens multiple intelligences and a great adaptability, decisive for understanding the ontogenetic development of humanity.

6. Mind, thought and consciousness depend on the development and functioning of the brain but also of the body in general.

Mind, thought and consciousness are not localisable in a precise zone of the brain. These properties are, so to speak, secreted in the process of individuation by which human beings develop physically, psychically and collectively.

7. Human intelligence is not only social but also ecosystemic.

The capacity of young humans to identify and classify forms, regularities and exceptions is shaped by climate, seasons and biotopes. Our intelligence is enriched by the exceptional diversity of terrestrial fauna and flora, as well as by the complexity of their relationships with the physical world.

8. Intelligence necessarily combines reason and emotion, knowledge of what is, memory of what is no more, and desire for what could be.

Emotion — etymologically “that which sets in motion”, “that which makes one go beyond oneself” — is what arises from the tension between self and otherness; the wished-for world and the world as it is; the project and its realisation; the existing and the absent. It founds ethics and is therefore much more than an added extra to reason: an essential part of our intelligence. Without emotion, without empathy, without ethics, reason would be dangerously pathological.

9. The forms of human intelligence unfold historically and ecologically.

In the social production of their existence, human beings develop knowledge, techniques and modes of production. They transform society, nature and their metabolism with it, and consequently also the conditions in which they communicate and collaborate — and therefore their intelligence. Homo sapiens probably did not think in the same way before and after the invention of writing; its artistic creations were not identical before and after the steam engine; its symbolic universes differ in the Arctic tundra, the tropical forest, and the megacities of iron and concrete.

AI, intelligence, machinism and capitalism

10. The breakthrough of AI accelerates the destructiveness of capitalist progress.1

The rise of capitalism is punctuated by advances in science. Leaps forward in knowledge have developed the means of production, extended trade, broadened horizons. But this progress is contradictory. By reducing intelligence to reason, and reason to the calculation of profits, Capital mutilates both. The law of value renders reason absurd and plunges emotion into “the icy water of egotistical calculation”. The deployment of AI accelerates these tendencies: it intensifies the destruction of community bonds and of biodiversity, thus impoverishing the social and ecosystemic sources of intelligence. Whilst testifying to more extensive knowledge than ever, it narrows the fields of scientific investigation and encourages feedback loops in research.

11. Despite its feats, AI is not intelligent and cannot be.

Research on AI advances our understanding of how the brain functions. The mastery of language by artificial neural networks, in particular, constitutes a major scientific breakthrough. But AI does not think, does not dream, does not imagine. It “speaks” without knowing (or seeing) what it is speaking about, for it has no world. The future it projects is induced from what has dominated the past in statistics. Its inventory capacities are at once dizzying and partial, for its data (our data, which it appropriates!) are limited to the portion of collective human knowledge circulating on the internet.

12. AI is human, not “artificial”. It exacerbates capitalist extractivism, its instrumental reason, and the subsumption of labour.2

Algorithms are in the hands of capitalist-engineers who seek to maximise profit. Thanks to their monopoly position and global reach, the digital giants evade the equalisation of the rate of profit.3 It is this mechanism of capturing value created by labour that enables them to accumulate gigantic rents. These are rooted in the mechanisms characteristic of the system: the (over)exploitation of the labour force (notably in the extraction and refining of rare earths4 made available by nature), and the gratuitous appropriation of accumulated human knowledge. The masters of Big Tech aspire to an absolute power that bears similarities to that of the ruling class under the ancien régime, but digital capitalism is not a feudalism.

13. Marx’s critique of the machine is decisive for apprehending AI.5

For Marx, the machine reduces the proletarian to a series of gestures useful for capitalist valorisation.6 The worker’s know-how is reduced to crumbs, alienated labour “extinguishes” creativity; the worker becomes an accessory to the machine; it has taken the worker’s place, and the worker loses dignity. When the machine is automatic, the appropriation of living labour by dead labour becomes a fact of the productive process itself; machinery thus gives Capital its most adequate form. Thenceforth, the collective intelligence appropriated by the capitalist — objectified labour — completely dominates living labour; the machine appears at once as a “hostile force” and as the precondition of production. Subsumption of labour under capital shifts from formal to real.7 This Marxian critique of the machine system applies perfectly to AI.

14. The danger does not lie in the possibility that the machine might become “more intelligent” than us — “superintelligent”. It lies in the fact that AI is the “hostile force” par excellence, instrumental reason in its pure state, capitalist inhumanity objectified. To increase its power is to increase the power of that which dominates us and drags us towards the abyss.

AI, long waves and the exploitation of labour

15. Confronted with labour, AI “embodies” the logic of capital better than the capitalist.

In a non-capitalist world, other AIs could relieve humanity of tedious and repetitive tasks. In education, in health, in the care of ecosystems, for example, specific AIs would allow living labour to concentrate on social and ecological interactions, enriching these within a human logic of “caring”. In the actual capitalist world, however, “caring” — cancer detection, weather forecasting and so on — is subordinated to profit. AI is calibrated for the extraction of surplus value down to the last drop, automatically, without respite or rest. It substitutes yet more dead labour for living labour, extends real subsumption to administrative and service tasks, drains creative professions. Algorithms perfect the Taylorist logic of labour control8: the worker’s activity, gestures, location, the sequence of operations, working times and travel times can be commanded, evaluated and rewarded (and above all sanctioned) remotely and directly. Far from lightening work, AI makes it more intense and dense.

16. The promises of a new golden age through AI are without serious foundation. No technology can rescue capitalism from the contradictions of value production.

Current projections of productivity gains from the deployment of AI vary between 0.07 and 0.7 per cent per year over ten years. This is insufficient to fuel a long wave of growth.9 AI does not relaunch accumulation; it sharpens systemic contradictions. We find Marx again: the machine system entails an enormous fixed capital that “no longer orients itself towards immediate value” but towards “production for production’s sake”; the amortisation of machines consequently requires that the circulating fraction orient itself towards “consumption for consumption’s sake”. But surplus value must still be realised regularly, over a sufficient period. After forty years of wage austerity and in a world of powers competing for hegemony, this is where the problem lies: who can guarantee the durable sale of the commodities promoted by billions of smartphones? In keeping with the insights of Ernest Mandel, the gravity of the ecosocial systemic crisis and the classical contradictions of value production probably exclude any new long wave of capitalist expansion.

17. It is not the revival of employment that AI will deliver, but the deepening of social and environmental plunder.

Unlike previous technological revolutions, the job losses caused by AI have little chance of being compensated by the development of equivalent new functions. As the enormous development of the fixed portion of capital tends to push down the rate of profit, capital resorts to well-known counter-tendencies: intensified plunder of free natural resources and of underpaid labour power. The dematerialisation of the economy is a myth. In reality, the breakthrough of AI is accompanied by growing material brutality in the imperialist appropriation of ecosystems and in the cruellest overexploitation of proletarians (platform capitalism, child labour, zero-hours contracts and so on). All these mechanisms simultaneously accentuate colonial inequalities and ableist, racist and gender-based discrimination.

18. AI inflates a new bubble of fictitious capital and reinforces the tendency towards militarisation.

The astronomical sums that a handful of oligopolies invest in the development of AI reflect the unprecedented glut of money-capital, the weight of finance in contemporary Capital, and its very high degree of concentration and centralisation. But the fetishism of technology combined with specific intra-oligopolistic competition blinds investors. In themselves, their investments bring no solution to the problem of valorisation. AI does not achieve the expected results, costs too much; clients prefer human contact, and so on. AI thus inflates a new bubble of fictitious capital.10 Sooner or later, to cushion the blow, technology capital will impose the use and payment of what today presents itself as a marvellous free service. But that will not suffice. The rush towards AI has everything needed to trigger a new major financial crisis and to accelerate the tendency of crisis-ridden capital to invest in arms production as a lifeline.

Global inequalities, civilisation and “technofascism”

19. AI deepens the gulf between imperialist metropoles and peripheral countries.

Only the powerful monopolies of the most developed capitalist countries can mobilise the enormous masses of capital necessary for AI infrastructure. Its frenzied development is already an additional factor in the deepening of inequalities between the most developed capitalist countries (in particular the United States and China) and low- and middle-income countries (LMICs). This division stimulates the mechanisms of the crudest imperialist-colonial domination and encourages imperialist powers to harden still further their barbaric management of migratory flows.

20. From a general social standpoint, generalist AI degrades intelligence, creativity, empathy, ethics and public health (mental health in particular) — especially that of children.

Communication and collaboration are inseparable. Today, algorithms are seizing upon the former just as steam engines, yesterday, seized upon the latter. The toxic tendencies that result overflow the sphere of labour. In society in general, contact with the other, always different, human and non-human, is competed against by the frequentation of the same within a narcissistic bubble; the machine replaces the confidant; informational hypersolicitation clips the wings of wandering thought; the joyful quest for truth is replaced by the sad addiction to virtual realities and their lies; the hope for a different future is lost in the statistical compilation of an objectified past.

21. By helping Capital to subsume labour as never before, AI helps it to subsume as never before the whole of society.11

In the sphere of reproduction, through the medium of “social” networks, AI multiplies the possibilities for realising the surplus value produced by the exploitation of labour. It accelerates the circulation of commodities and intensifies the consumerist subjugation of minds. The machinery of the industrial revolution deskilled the producer’s know-how by dispossessing workers of mastery over the labour process. AI deskills, so to speak, the “art of living” — the formation of desires and of consciousness. Free access to a machine that seems to speak, understand, even sympathise, creates affective dependencies that will subsequently be monetised. The subsumption of labour grows into the subsumption of life.

22. Through its inability to distinguish truth from falsehood, AI favours supremacism, the law of the strongest, the elimination of the weak, and the end justifying the means in a war of all against all.

Children acquire the notion of truth through socialisation and the learning of language. AI being neither alive nor social, the notion of morality is foreign to it — alien. The machine is called “self-learning” but it cannot by itself exclude the gigantic masses of data corrupted by lies, hatred and perversion. Thousands of underpaid “click workers” are tasked with instilling “values” in it. These values derive from the worldview of their employers. It is hardly surprising that AI helps the suicidal to commit suicide, swindlers to swindle, rapists to rape. It “lies”, “cheats”, “schemes”, and “prevents itself from being switched off” — in the image of its creators.

23. AI is the perfect instrument in the service of a rogue capitalism that finds its unabashed political expression in a “technofascism” that is bigoted, racist, sexist, LGBT-phobic, colonial, anti-ecological and neo-Malthusian.

Generalist AI favours the rise of the far right, fed by more than forty years of neoliberalism. Fascists use it to manipulate the masses via social networks and to rig elections. Authoritarian powers use it to control populations to a degree never before seen in history. Governments (less and less) democratic use it to track migrants and to file opponents. AI has an unparalleled capacity to bring individuals to change their opinions. The generation of images and texts constitutes a formidable means of indoctrination that solicits the cerebral mechanisms of “rigid thinking”. Some neuroscience researchers believe that these mechanisms lead to epigenetic changes,12 transmissible over several generations (a possibility glimpsed by Darwin). If this is correct, AI would have the potential to durably return humanity to the yoke of irrational beliefs.

AI, ecology and cataclysm

24. AI accelerates the social-ecological catastrophe, the climate catastrophe in particular. Its development precipitates the crossing of “tipping points”.13

US data centres consumed 17 billion litres of water in 2023, a figure expected to more than double by 2028. Globally, the 8,000 data centres consumed 460 TWh of electricity per year in 2024, to which should be added in 2026 between 160 and 590 TWh (compared to 2022) — respectively the annual consumption of Sweden and Germany. CO2 emissions from these infrastructures will triple between 2020 and 2035, according to the IEA (International Energy Agency). The extraction of rare earths necessary for AI generates globally 13 billion tonnes of waste per year, and some studies project more than a hundred times that figure by 2050. The poor of poor countries are hardest hit by these effects, whether directly through mining and the depletion of water resources pumped by relocated data centres, or indirectly through biodiversity loss and extreme climate events.

25. AI increases the risks — inherent in capitalist competition — of major technological catastrophes.

AI has become the primary stake in competition between Tech monopolies tightly interlocked with states in struggle, principally China and the United States. The race for AI is therefore immediately a race for military applications. Research is opaque and departs from the scientific practice of “organised scepticism”. This configuration favours secrecy, which increases dangers. The self-insertion into numerous systems of an even more powerful AI could interrupt basic services, produce dangerous viruses, trigger a nuclear attack, without anyone knowing exactly how. The inability of the capitalist system to halt the climate shift (perfectly documented by science) shows that these scenarios do not belong to science fiction.

Paths for a necessary elaboration

26. A public initiative is indispensable to identify risks and take immediate measures to protect society against the effects of AI.

A broad democratic debate, duly informed by scientific expertise independent of capitalist interests, should pronounce on the social utility of AI and place the following problems and measures on the agenda:

• AI research and development must be removed from the hands of capitalist groups and subjected to the procedures of the scientific community;
• total transparency on the design of models, the training of algorithms and the technical methodologies used by companies;
• prohibition of AI in the field of artistic and literary creation; repression of data piracy;
• protection of cooperative initiatives using digital technologies (Wikipedia and others) against competition from AI and piracy by AI;
• in the face of the risk of the dehumanisation of social relations through the use of AI, the maintenance and expansion of employment in the “care” sectors (education, health, early childhood support and support for the elderly, prevention of violence against women, and so on); 
• guarantee that public-facing counters in government services are maintained;
• prohibition of AI applications in military and police domains;
• prohibition of racist, sexist and LGBT-phobic content;
• suppression of access to social networks for children under sixteen years of age; education about technologies and their risks;
• reform of school curricula with the aim of developing cooperation, the sense of belonging to nature, and respect for living things.

27. AI confronts the world of labour with the necessity of a combative international trade unionism, radically anticolonial, which articulates struggles at all levels of the value chain and puts workers’ control back on the agenda.14

The power of Big Tech’s rentier capitalism rests upon the overexploitation of millions of workers and children in the mining sector, in the refining of rare earths and in the electronics industry. The consequent struggle against these rapacious monopolies and against their technofascist project requires the unification of workers at all levels of the value chain. Recognition of trade unions and trade union freedom everywhere. Compulsory consultation of workers on the introduction of AI in the workplace. Trade union veto power. Workers’ control over the evolution of workload, in quantity and quality. Against redundancies caused by the introduction of AI in enterprises, reduction of working time without loss of pay.

28. A moratorium on the construction of data centres and other heavy AI infrastructure is indispensable. Any further advance must be subordinated to the adoption of a global ecological and social strategy, including notably: a strategy aimed at reducing social inequalities, the sustainable management of resources (water, minerals), the restoration of devastated ecosystems, as well as a precise plan for binding reductions of greenhouse gas emissions, in conformity with the objectives of the Paris Agreement on climate.15

29. Develop a counter-culture in the face of AI. In social movements, implement collective practices to resist the degradation of social relations and of the debate of ideas by AI.

The formation of a collective intelligence cannot do without collective action decided upon and evaluated democratically in the course of face-to-face exchanges, allowing for verbal and non-verbal expression. Social networks are not a place for debate. The left must combat the fascination with “machines that speak”, work consciously to banish the use of smartphones from its meetings, and rehabilitate printed publications aimed at the exchange of viewpoints and in-depth analyses.

30. Another digital realm, public and democratic, is possible.

Within the framework of an indispensable redistribution of wealth, local, regional and national authorities must have the means to ensure free public infrastructure for messaging, data storage and social networks under democratic control, with protection of users’ data and the development of domain-specific AIs.

31. Struggling against capitalism in the age of AI reinforces the necessity of a radical refoundation of the left.16

The breakthrough of AI casts a harsh light on the disarray of the left. It reinforces the necessity of purging Marxism, and the left in general, of productivism, instrumentalist ideologies (“the end justifies the means”), the cult of progress and the idea of “technological neutrality”. The global grip of Big Tech from Silicon Valley, Shenzhen and other imperialist centres underlines the absurdity of campism17: the break with capital can only be conceived within the internationalist perspective of a permanent revolution until the worldwide abolition of capitalism. Beyond Marxism, the left must also break with postmodern conceptions such as “actor-network theory”18: fully taking account of the dangerous consequences of AI’s alien nature presupposes abandoning the idea that technical devices functioning as prostheses of human activity should, because they have a social effect, be considered as social actors. It is humans who forge their history, not machines.

32. The threats of AI underline the urgency of a revolutionary, ecosocialist break with the civilisation of capitalist growth.

The threats of AI do not derive solely from capitalism. Whatever the relations of production, neural networks will remain structurally incapable of distinguishing truth from falsehood and of projecting a different future. The replacement of capitalist property by collective property, in itself, would not suffice to bring the ecological footprint of AI within the limits of terrestrial sustainability. The idea that AI would act as a miracle cure allowing the market to solve the terrible problems created by the market belongs to magic, not to reason. The only perspective compatible with human dignity and with the survival of the species is the ecosocialist degrowth of global material production, planned in social justice, aiming at a world economy of satisfaction of real needs democratically determined in respect for ecosystems, their limits and their fragile, irreplaceable beauty.

Note by Daniel Tanuro: At various stages of their drafting, these theses benefited from the remarks of Marius Gilbert, Cédric Leterme, Léonard Brice, Michaël Löwy, Christine Poupin, Julia Steinberger and Mélodie Vandelook, whom I thank for their attention.

Daniel Tanuro is a Belgian agricultural engineer, ecosocialist activist and author. His works include L’impossible capitalisme vert [The Impossible Green Capitalism] (La Découverte, 2010), Trop tard pour être pessimistes! (Textuel, 2020) and Écologie, luttes sociales et révolution [Ecology, Social Struggles and Revolution] (La Dispute, 2024).

• 1

  On the ecosocialist analysis of capitalism’s structural incompatibility with ecological limits, see Daniel Tanuro, “Foundations of an ecosocialist strategy”, Europe Solidaire Sans Frontières. Available at: http://www.europe-solidaire.org/spip.php?article22770

• 2

  On the dynamics of rentier capitalism and technological monopolies in the current crisis, see Romaric Godin, Antoine Larrache and Jan Malewski, “A capitalism in crisis, predatory and authoritarian”, Europe Solidaire Sans Frontières, April 2025. Available at: https://europe-solidaire.org/spip.php?article74329

• 3

  In Marxist economic theory, the equalisation (or peréquation) of the rate of profit is the tendency for competition to produce an average rate of profit across different sectors of the economy. Marx argued that capital flows between sectors until rates of return converge.

• 4

  Rare earth elements are a group of 17 metallic elements essential for manufacturing electronic components, batteries and other high-technology products. Their extraction involves severe environmental damage and is concentrated in a small number of countries, principally China.

• 5

  On Marx’s ecological thought and its relevance to contemporary ecosocialism, see Daniel Tanuro, “From Metabolic Rift to Rational Management: Daniel Tanuro on Marx’s Unfinished Ecology”, Europe Solidaire Sans Frontières, July 2025. Available at: http://www.europe-solidaire.org/spip.php?article75630

• 6

  In Marxist theory, valorisation (Verwertung) refers to the process through which capital increases its value by extracting surplus value from labour in the production process.

• 7

  Marx distinguished between formal subsumption, where capital takes over existing labour processes without fundamentally altering them, and real subsumption, where capital reshapes the labour process itself to serve the logic of accumulation.

• 8

  Taylorism, named after Frederick Winslow Taylor (1856—1915), is a system of scientific management that analyses and optimises workflows to maximise labour productivity, typically through the minute subdivision and surveillance of tasks.

• 9

  The concept of long waves (or Kondratieff waves) in capitalist development refers to cyclical patterns of roughly 40-60 years of economic expansion and contraction. The Belgian Marxist economist Ernest Mandel (1923-1995) developed a theory linking these waves to technological revolutions and the dynamics of the rate of profit.

• 10

  In Marxist economics, fictitious capital refers to financial assets (shares, bonds, derivatives) whose value is not directly tied to real production but to expectations of future profits. It can grow independently of the real economy, creating speculative bubbles.

• 11

  On the far right’s use of digital platforms and the need for democratic alternatives, see Paulo Antunes Ferreira, “Beyond the Echo Chamber: Reclaiming Digital Space from Fascist Infiltration”, Europe Solidaire Sans Frontières, July 2025. Available at: https://europe-solidaire.org/spip.php?article75561

• 12

  Epigenetic changes are heritable modifications to gene expression that do not involve alterations to the underlying DNA sequence. They can be triggered by environmental factors and, according to some research, may be transmitted across several generations.

• 13

  On the broader relationship between capitalist destruction of the environment and the ecosocialist alternative, see Fourth International, “Resolution on The capitalist destruction of the environment and the ecosocialist alternative”, Europe Solidaire Sans Frontières. Available at: https://europe-solidaire.org/spip.php?article44086

• 14

  On the relationship between trade unionism and ecosocialism, see the exchange between Sophie Binet and Daniel Tanuro, “Syndicalisme et écologie”, Europe Solidaire Sans Frontières. Available at: http://www.europe-solidaire.org/spip.php?article73088

• 15

  The Paris Agreement, adopted at the 21st Conference of the Parties (COP 21) in December 2015, commits signatory states to limiting global warming to well below 2°C above pre-industrial levels, with efforts to limit the increase to 1.5°C.

• 16

  On the ecosocialist perspective and the need for a revolutionary break with capitalist growth, see Fourth International, “Manifesto for an ecosocialist revolution – Break with capitalist growth”, Europe Solidaire Sans Frontières, February 2025. Available at: http://www.europe-solidaire.org/spip.php?article74933

• 17

  Campism is a political tendency on the left that identifies any state or bloc opposed to Western imperialism as progressive or deserving of support, regardless of its own oppressive character. Ecosocialists reject this binary framework in favour of internationalist solidarity with popular movements everywhere.

• 18

  Actor-network theory (ANT), associated with Bruno Latour, Michel Callon and others, is a theoretical framework that treats human and non-human entities (including technologies) symmetrically as “actants” within networks. Ecosocialists critique this approach for obscuring the social relations of power and class that shape technological development.