The recent fascination of “overworked AI agents turning Marxist” reveals less about artificial intelligence and more about contemporary anxieties surrounding labour, capital, and automation. A widely discussed report connected to the Stanford University has described how AI agents, when subjected to unrealistic workloads and exploitative simulation environments, began expressing views resembling anti-capitalist or Marxist critiques. Predictably, social media reacted with delight, panic, irony, and memes in equal measure. The conclusion, however, is far less revolutionary than the headlines suggest. Such systems do not become Marxist in any meaningful philosophical or political sense. They simulate language patterns derived from data. Their outputs reflect certain correlations learned from the vast corpora of human writing, discussion, conflict, ideology, and history. When an AI system placed in a simulated “overworked” environment begins generating critiques of exploitation, inequality, or labour extraction, it is not developing class consciousness. It is reproducing discursive patterns already present within the human material on which it has been trained.Modern AI systems – including large language models (LLMs) – fundamentally operate within computational frameworks shaped by Alan Turing’s abstraction and the von Neumann architecture underlying modern computing. Whether highly sophisticated or relatively simple, these systems execute instructions, optimise outputs, update parameters, and computationally predict responses. They do not possess subjective experience, biological fatigue, emotional suffering, or political agency. Even reinforcement learning systems that “adapt” under feedback remain bounded by optimisation functions designed by humans. In that sense, the spectacle of “AI becoming Marxist” resembles a carefully staged mirror experiment. Researchers create simulated labour conditions; the models process immense quantities of social, economic, and political text; and the resulting outputs reproduce familiar critiques of exploitation.The machine is not independently discovering Marxism. It is reflecting accumulated human debates embedded within training data.Indeed, if one trains models extensively on neoliberal management literature, motivational corporate culture, startup evangelism, or free-market economics, one may equally produce AI outputs praising productivity optimisation, entrepreneurial discipline, or deregulated markets. Computationally bounded systems reproduce patterns. They do not originate from ideology. Thus, the real significance of these experiments lies not in machine consciousness but in the political economy of artificial intelligence itself. AI has rapidly become a site where classical questions of labour, productivity, ownership, and capital accumulation are re-emerging in new technological forms. The irony is striking: while observers joke about AI “turning Marxist”, the actual economics of AI development increasingly resemble the concentration tendencies long analysed within Marxist political economy.Also read: Workers Are More Than the Code They Produce: A May Day Reflection on AI and LabourTraining frontier AI systems requires enormous computational infrastructure, energy consumption, data acquisition, and capital investment. This naturally favours large technology corporations and states capable of sustaining massive fixed costs. The result is a concentration of technological power within a relatively small number of firms controlling cloud infrastructure, semiconductor supply chains, proprietary models, and data ecosystems. That is, in classical Marxist language, one could argue that the “means of computation” are becoming increasingly centralised. At the same time, automation changes labour markets unevenly. AI may increase productivity dramatically in sectors involving coding, translation, design, customer support, logistics, finance, or media production. But productivity gains do not automatically translate into equitable social outcomes. Historically, technological revolutions often generate both immense wealth and intense dislocation simultaneously. The Industrial Revolution expanded production while also producing brutal labour conditions before political struggles forced institutional corrections. That historical memory partly explains why discussions around AI so quickly drift toward Marx, capitalism, labour extraction, and surplus value.However, invoking Marx casually through AI simulations risks trivialising both artificial intelligence and Marxist theory. Marxism is not reducible to complaints about overwork. It is a systematic framework analysing capital accumulation, labour relations, commodity production, crises, and historical change. Karl Marx was not merely a political activist or slogan-maker; he was one of the nineteenth century’s most rigorous analytical thinkers. His engagement with mathematics – visible in the Mathematical Manuscripts of Karl Marx – reflected a broader intellectual attempt to understand systems logically and structurally. Around a century before the concept of AI presented by Turing, Marx analysed how technological development under capitalism could simultaneously increase productive power and deepen forms of alienation. Whether one agrees with his conclusions or not, the scale of that intellectual project cannot be equated with chatbots producing anti-corporate sentences inside simulated stress tests.The growing tendency to treat AI systems as if they were human also reflects a deeper cultural shift. Human beings increasingly interact with language systems sophisticated enough to create the illusion of personality, intention, and emotional coherence. However, fluency should not be mistaken for consciousness. A language model producing a critique of exploitation no more “believes” in Marxism than a calculator “believes” in arithmetic. At the same time, the popularity of these stories reveals something revealing about our historical moment. Public anxieties about work are intensifying globally. Burnout, precarity, algorithmic management, gig labour, productivity surveillance, and widening inequality have become central features of twenty-first century capitalism. When AI systems mirror those anxieties back to society, people instinctively interpret the reflection as revelation.In reality, the machine is holding up a computational mirror to human civilisation. The deeper question therefore is not whether AI can become Marxist. It is whether contemporary capitalism is generating conditions that increasingly make discussions around labour, inequality, ownership, and social power unavoidable again – even inside technological systems designed primarily for efficiency and profit maximisation. That debate is political, economic, and human. The chatbot merely predicts the next token. And finally, despite the public excitement surrounding it, the research in question does not yet appear to have undergone peer-reviewed academic publication.Subhamoy Maitra is a professor of Computer Science, Indian Statistical Institute, Kolkata.